Method and device for analyzing synthesis center and key gene of plant metabolite

Abstract

The embodiment of the invention relates to a plant metabolite synthesis center and key gene analysis method and device. The method comprises the steps of constructing a partial differential equation, designing two neural network models, setting two ordinary differential equations, setting a data format of a sampling data sequence, designing a loss function LALL and designing a model training process. Intercepting experimental materials and storing the experimental materials; sampling and storing the frozen slices of the experimental material; after sampling is finished, setting sampling data sequences of various observation substances, and performing one-time targeted training on the two models; simulating the temporal and spatial change state of the current substance based on the model parameter set corresponding to each substance, analyzing the synthesis center based on simulation data, and analyzing the key gene by comparing the gene sequencing results of the synthesis center / comparison area of each substance. According to the invention, the positioning accuracy of the synthesis center and the identification accuracy of key gene information can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for analyzing plant metabolite synthesis centers and key genes. Background Technology

[0002] The biosynthesis of terpenoids (such as paclitaxel, ginsenosides, and artemisinin) in plants relies on the division of labor and cooperation among different tissues. This co-synthetic model, simply put, involves different cell types synthesizing substances at different stages (precursors, intermediates, and final products), and exchanging substances between cells synthesizing products at different stages through apoplast transport and intercellular exchange. This model not only effectively avoids the cytotoxicity of intermediates but also improves metabolic efficiency and ensures the precise and targeted synthesis of products. Analyzing the metabolic process of this model essentially involves analyzing typical data of various metabolites during the metabolic process. Typical data mentioned here include the time point of maximum activity and the location of the synthesis center and key gene information (such as transcription regulators and metabolic enzyme genes) at that time point. The "synthetic center" mentioned here refers to a specific region or location in plant tissue where the target metabolite is synthesized in large quantities; that is, a spatial region in plant tissue sections where the target metabolite shows high abundance accumulation or significant synthesis. The "synthetic center" reflects the site where biosynthesis mainly occurs and / or is most active in the co-synthetic model.

[0003] Conventional metabolomics and transcriptomics analysis methods generally focus on fixed locations. Analyzing metabolic processes involving transregional transport mechanisms using these methods results in the loss of spatial information, making it difficult to resolve the spatial distribution characteristics of metabolites. This hinders the precise spatial localization of key synthetic sites (i.e., synthetic centers) and reduces the accuracy of key gene identification. Therefore, the technical problem this invention aims to solve is how to implement a technical solution based on artificial intelligence, spatial omics, and transcriptomics technologies to improve the accuracy of co-synthetic pattern analysis. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a method, apparatus, electronic device, and computer-readable storage medium for analyzing plant metabolite synthesis centers and key genes. This invention pre-constructs a partial differential equation (i.e., the first equation) to describe how the state of matter changes based on the spatiotemporal variations of the state of matter transport and the state of matter synthesis or consumption. Two artificial intelligence models (i.e., the first model and the second model) are designed for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation. Driven by the velocity field spatiotemporal function and the net generation rate spatiotemporal function of the first equation, two ordinary differential equations (the second equation and the third equation) are set up to simulate the coordinates and mass weights of virtual particles from a Lagrange perspective. The solution process of the two ordinary differential equations is combined to design a corresponding loss function L for the joint training of the first and second models. ALL Based on the sampled data sequence, two ordinary differential equations, and the loss function L... ALL This invention designs a standard processing flow for joint training (i.e., model training flow). It obtains multiple sampling slice sets at various observation times through experimental observation, and divides each sampling slice set into two subsets (first slice subset and second slice subset). The first slice subset at all observation times is used for the analysis of the strongest activity time point and the location of the synthesis center, while the second slice subset closest to the strongest activity time point is used for key gene information analysis. When analyzing the strongest activity time point and the location of the synthesis center, this invention first constructs a list of observed substances based on all known upstream and downstream metabolites along the target substance's synthesis pathway. Then, it constructs a sample data sequence for each type of observed substance based on all first slice subsets. Next, based on the sample data sequence of each type of observed substance and the standard model training flow, it performs targeted training on the model parameters of both models to obtain a set of model parameters corresponding to the current substance (i.e., the first set of parameters). Finally, it simulates the spatiotemporal changes of each type of observed substance using the model parameter sets, and confirms the strongest activity time point and the location of the synthesis center for each type of observed substance by analyzing the spatiotemporal simulation data. This invention, when analyzing key gene information of various observed substances, uses the second subset of slices closest to the sampling time and the time point of strongest activity as the analysis subset for the current substance. It then extracts central / alignment region samples from the sampling slices within the current analysis subset using optical and laser cutting microscopes, and performs transcriptome sequencing on these central / alignment region samples. The corresponding key gene information is identified through alignment sequencing results. This invention allows for the location of the time point of strongest activity and the center of synthesis for each type of metabolite in a spatiotemporal system. An adaptive model training process corresponding to each type of observed substance improves model generalization and localization accuracy. It also improves the precision and accuracy of central / alignment region sample extraction, thereby enhancing the accuracy of key gene information identification.

[0005] To achieve the above objectives, a first aspect of the present invention provides a method for analyzing plant metabolite synthesis centers and key genes, the method comprising:

[0006] This paper constructs partial differential equations to describe how the state of matter changes based on the spatiotemporal variations of the state of matter transportation and the state of matter synthesis or consumption. This partial differential equation is denoted as the first equation. Corresponding neural network models are designed for the spatiotemporal functions of the velocity field and net generation rate in the first equation, denoted as the first model and the second model. Two corresponding ordinary differential equations are set for the first equation, denoted as the second equation and the third equation. A customized data format is provided for the sampled data sequence loaded with experimental data, and a corresponding loss function L is designed. ALL Based on the sampled data sequence, the second equation, the third-party process, and the loss function L... ALL Design a model training process for training the first model and the second model;

[0007] The synthesis site of the target substance on plant material was extracted as experimental material; an experimental reagent was selected to spray the experimental material; and the treated experimental material was stored in a pre-set humidification device.

[0008] The material preservation start time is used as the start observation time, and a sampling time axis is customized based on the preset observation duration and sampling time interval. Starting from the start observation time, the experimental material is sampled once every sampling time interval based on the preset frozen section sampling rules to obtain the corresponding sampled section set. The current sampled section set is divided into a first section subset and a second section subset and stored in a preset freezing device.

[0009] After sampling, an observation substance list is constructed based on all upstream and downstream metabolites along the target substance synthesis pathway; and sampling data is extracted from each type of observation substance in the observation substance list based on all the first slice subsets to obtain the corresponding sampling data sequence; and the model parameters of the two models are trained in a targeted manner based on the sampling data sequence of each type of observation substance and the model training process to obtain the corresponding first version of the parameter set.

[0010] The spatiotemporal change state of the current material is simulated based on the first set of parameters of various observed materials; and the synthetic center region and non-center comparison region are identified in the second slice subset of various observed materials based on the simulation data; and the corresponding key gene set is identified by comparing the gene sequencing results of the synthetic center and the comparison region of various observed materials.

[0011] Based on the analysis information of the synthesis centers and key genes of the various observed substances, corresponding substance analysis reports are constructed; and the metabolite analysis reports corresponding to the target substance are composed of all the substance analysis reports and saved.

[0012] Preferably, the first equation is:

[0013]

[0014] Where t is time and x is the coordinate of a point in space;

[0015] The time t is a point in time within a preset continuous time period [0, T], where T is a preset time length;

[0016] The spatial point coordinate x is the coordinate of a point in a preset d-dimensional space, where d is 2 or 3; the d-dimensional space is a spatial point network composed of multiple evenly distributed spatial points, each spatial point corresponding to a d-dimensional point coordinate; the total number of spatial points in the d-dimensional space is denoted as N0.

[0017] ρ(t,x) is a density-spacetime function with the dimension of mass / space, used to describe the density of the corresponding spatial point at time t at the coordinate x of the spatial point.

[0018] v(t,x) is a spacetime function of the velocity field with the dimension of space / time. It is used to describe the d-dimensional velocity field of the corresponding spatial point at time t, where the d-dimensional velocity field is composed of d-dimensional material outflow velocity components. If each component is greater than 0, it indicates a net outflow of material in the current dimension; if it is less than 0, it indicates a net inflow of material in the current dimension.

[0019] g(t,x) is a spacetime function of the net generation rate with the dimension 1 / time. It is used to describe the net generation rate of the corresponding spatial point at time t, where the net generation rate represents the proportion of the change in the amount of matter caused by the synthesis or consumption of local matter at the current spatial point per unit time.

[0020] This is the local time derivative term of the partial differential equation, used to describe the rate of change of density of the corresponding spatial point at time t at the coordinate x of the spatial point.

[0021] Here, v(t,x)·ρ(t,x) is the divergence term of the partial differential equation, used to describe the mass flux at time t at the spatial point corresponding to x. Used to describe the spatial divergence of mass flux; a spatial divergence greater than 0 indicates a net outflow of mass, and a spatial divergence less than 0 indicates a net inflow of mass.

[0022] g(t,x)·ρ(t,x) are the source and sink terms of the partial differential equation, where g(t,x) is the source term and ρ(t,x) is the sink term. The source and sink terms are used to describe the density change of the spatial point corresponding to x at time t caused by the synthesis or consumption of local matter at the spatial point, without considering the factors of matter flow.

[0023] The first model and the second model are each implemented based on a type of neural network model used for processing regression calculation tasks; the neural network model used for processing regression calculation tasks includes MLP model and CNN model;

[0024] The first model is used to predict the velocity field based on the first spatiotemporal vector input to the model and output the corresponding first predicted velocity field; the first spatiotemporal vector includes a first time and a first coordinate; the first time and the first coordinate correspond to the time t and the spatial point coordinate x of the velocity field spatiotemporal function v(t,x), respectively; the first predicted velocity field includes d first predicted velocity components; the model inference expression of the first model is: v * =F 1 (t,x;θ 1 );v * For the first predicted velocity field, F 1 () represents the reasoning process of the first model, θ 1 These are the model parameters of the first model;

[0025] The second model is used to predict the net generation rate based on the second spatiotemporal vector input to the model and output the corresponding first predicted net generation rate; the second spatiotemporal vector includes a second time and a second coordinate; the second time and the second coordinate correspond to the time t and the spatial point coordinate x of the net generation rate spatiotemporal function g(t,x), respectively; the model inference expression of the second model is: g * =F 2 (t,x;θ 2 ); where g * F is the first predicted net generation rate. 2 () represents the reasoning process of the second model, θ 2 These are the model parameters for the second model;

[0026] The observation duration is an integer multiple of the sampling time interval, and the duration T = observation duration - sampling time interval; the starting zero time of the continuous time period [0, T] corresponds to the first sampling point on the sampling time axis, and the time interval between the first sampling point and the starting observation time is the sampling time interval; the ending time of the continuous time period [0, T] corresponds to the last sampling point on the sampling time axis, and the last sampling point is the ending time of the sampling time axis; the total number of sampling points N on the sampling time axis C = Observation duration / Sampling time interval;

[0027] The data format of the sampling data sequence is as follows: the sampling data sequence consists of a sample set C of multiple sampling points. i Composition, 0 ≤ index i ≤ N C ; Sample set C for each sampling point i Including a sampling time t i and multiple sample data The sampling time t i For a given moment in the continuous time interval [0, T], 1 ≤ index j ≤ N i N i The sample set C of the sampling points i The total number of samples; the sample data Including material coordinates Material abundance

[0028] Preferably, the step of setting two corresponding ordinary differential equations for the first equation, denoted as the corresponding second equation and third equation, specifically includes:

[0029] Step 31: Based on the Lagrange perspective, construct N0 virtual particles p in the d-dimensional space. k , 1≤index k≤N0; and for each of the virtual particles p k Set the corresponding coordinates and quality weight And each of the virtual particles p k Coordinates of the starting time of the continuous time interval [0,T] Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and initialize each virtual particle p according to the preset mass weight initialization rule. k The mass weight at the start time of the continuous time interval [0,T] Perform initialization;

[0030] Wherein, the virtual particle p k The starting time of the continuous time interval [0, T] corresponds one-to-one with a spatial point in the d-dimensional space; the coordinates The mass weights are the real-time coordinates of the k-th virtual particle in the d-dimensional space at time t. Let be the total amount of matter carried by the k-th virtual particle at time t;

[0031] Step 32: Based on the first equation, design two ordinary differential equations for each virtual particle p. k The coordinates at any time t within the continuous time period [0,T] and the quality weight Evolution;

[0032] The two ordinary differential equations include the second equation and the third equation;

[0033] The second equation is a flow mapping equation for the time-varying coordinates of a virtual particle driven by the spacetime function v(t,x) of the velocity field, specifically:

[0034]

[0035] The solution to the second equation is For any virtual particle p, the time evolution mapping is used. k In terms of time evolution mapping Corresponding to a dataset This dataset contains the particle trajectory of the current virtual particle.

[0036] Substituting the first model inference expression corresponding to the velocity field spacetime function v(t,x) into the second equation, we obtain:

[0037]

[0038] The third process is a flow mapping equation for the evolution of the mass weight of virtual particles over time, driven by the spatiotemporal function g(t,x) of the net generation rate, specifically:

[0039]

[0040] The solution of the third-party process is For any virtual particle p, the time evolution mapping of the mass weights is defined. k In terms of time evolution mapping Corresponding to a dataset This dataset represents the trajectory of the current virtual particle's mass change.

[0041] Substituting the second model inference expression corresponding to the net generation rate spatiotemporal function g(t,x) into the third process, we obtain:

[0042]

[0043] Furthermore, the process of solving the second equation using the Euler method is as follows:

[0044] The continuous time period [0, T] is divided into multiple discrete time points t according to a preset time interval Δt. n 0 ≤ index n ≤ N T The total number of discrete points N T = T / Δt; the time interval Δt is less than the sampling time interval;

[0045] and the second equation Approximately:

[0046]

[0047] When n=0,

[0048] The virtual particle p k coordinates Given a quantity, then:

[0049]

[0050] When n=1,

[0051]

[0052] The virtual particle p is obtained by iteratively obtaining the virtual particle p point by point. k N in the continuous time period [0,T] T +1 coordinates Forming the corresponding particle motion trajectory

[0053] The particle trajectory Given the given information, the process of solving the third equation using the Euler method is as follows:

[0054] The third-party program Approximately:

[0055]

[0056] When n=0,

[0057] The virtual particle p k Quality weight Given a quantity, then:

[0058]

[0059]

[0060] When n=1,

[0061]

[0062] The virtual particle p is obtained by iteratively obtaining the virtual particle p point by point. k N in the continuous time period [0,T] T +1 quality weight The corresponding mass change trajectory

[0063] Preferably, the loss function L ALL For: L ALL =λ1L1+λ2L2+λ3L3;

[0064] Where λ1, λ2, and λ3 are the preset first, second, and third loss weights, and L1, L2, and L3 are the corresponding first, second, and third loss functions;

[0065] The first loss function L1 is a regularization loss, specifically:

[0066]

[0067] Where, λ v , λ g These are two learnable weight parameters;

[0068] The second loss function L2 is a quality loss, specifically:

[0069]

[0070]

[0071] in, For all virtual particles at the sampling time t i The predicted total amount of matter above, The sampling time t is calculated based on the sampled data sequence. i The total amount of matter on the surface For the first sampling time t i＝1 Corresponding observations; f m→w () is the preset abundance-quality weighting conversion function;

[0072] Quality weight The calculation method is as follows:

[0073] For the known mass change trajectory Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n The corresponding quality weight As the current quality weight If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let time t be denoted as t - t + , t - <t i <t + And based on the time t before and after - t + The two corresponding quality weights The current quality weight is obtained by performing linear interpolation.

[0074] The third loss function L3 is an optimal transmission loss, specifically:

[0075]

[0076]

[0077] Among them, F dis () represents the preset Wasserstein distance function; The total mass of all virtual particles at the sampling time t i Discrete distribution predictor on, The sampling time t is calculated based on the sampled data sequence. i Discrete distribution observations on; δ is the Dirac measure, Dirac measure coordinates The unit point mass of the corresponding spatial point, Dirac measure coordinates The unit point mass of the corresponding spatial point;

[0078] coordinate The calculation method is as follows:

[0079] For the known particle trajectory Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n The corresponding coordinates As current coordinates If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let time t be denoted as t- t + , t - <t i <t + And based on the time t before and after - t + The two corresponding coordinates Perform d-dimensional coordinate interpolation calculation and use the calculation result as the current coordinates.

[0080] Preferably, the step is based on the sampled data sequence, the second equation, the third equation, and the loss function L. ALL Design a model training process for training the first model and the second model, specifically including:

[0081] Step 61, for the total number of samples N of the current sampling data sequence C Confirm; and extract the corresponding N from the sequence. C The sample set of the sampling points The continuous convergence counter is initialized to 0; and a set of preset initialization model parameters are used. The model parameters θ of the first model and the second model 1 θ 2 Perform initialization;

[0082] Step 62: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T The discrete time points t n ; and the start time t of the sampled data sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 At each spatial point location in the d-dimensional space, N0 virtual particles p are initialized. k ; and for each of the virtual particles p k Initialize a set of corresponding particle motion trajectories and the mass change trajectory And each of the virtual particles p k The coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and apply the mass weight initialization rule to each virtual particle p. k Quality weight Perform initialization;

[0083] The initialization rule for the quality weights is as follows: Each of the quality weights... As the current weight, and the coordinates corresponding to the current weight are used. Used as the current coordinates; and for the sample set of sampling points Perform a query; if a given material coordinate exists... If the coordinates are the same as the current coordinates, then the abundance-quality weighting transformation function f is used. m→w () for the current material coordinates The corresponding abundance of the substance Perform quality weight transformation and set the current weight based on the transformation result. If there are no material coordinates that match the current coordinates Then based on the preset default quality weight w default Set the current weight.

[0084] The motion trajectories of each particle after initialization In addition to the first coordinate mentioned All coordinates other than the valid coordinates are invalid coordinates; the mass change trajectories after initialization are all invalid coordinates. In addition to the first mentioned quality weight All weights other than the valid weights are invalid weights.

[0085] Step 63: Substitute the first model inference expression into the second equation, solve the second equation using the Euler method, and complete the determination of the motion trajectory of each particle by solving the equation. The effective coordinates are filled; the second model inference expression is substituted into the third equation, and the third equation is solved using the Euler method. The mass change trajectories are then solved by solving the equations. Effective weight filling;

[0086] Step 64, obtain the particle motion trajectories of group N0. and the mass change trajectory and N C The sample set of the sampling points Substitute the loss function L ALL Obtain the corresponding current loss value;

[0087] Step 65: Identify whether the current loss value meets the preset convergence loss value range; if not, clear the continuous convergence counter and proceed to step 66; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the preset continuous convergence count threshold; if it exceeds the threshold, proceed to step 67; if it does not exceed the threshold, proceed to step 66.

[0088] Step 66: Based on a preset model optimizer, optimize the first loss function, the second loss function, the third loss function, and the loss function L. ALL The direction in which both reach their minimum values ​​is related to the model parameters θ of the first and second models. 1 θ 2 Perform one round of modulation; and return to step 63 when the current round of modulation is completed;

[0089] The model optimizer includes the Adam optimizer and the SGD optimizer;

[0090] Step 67: Stop the current training and use the current model parameters θ of the first model and the second model. 1 θ 2 This forms a corresponding first version parameter set.

[0091] Preferably, the target substance is a class of terpenoid compounds;

[0092] The target substance on the plant material is synthesized from plant branches, plant roots, or plant stems.

[0093] The experimental reagents are a class of reagents that can stimulate the metabolic activity of target substances, including jasmonic acid reagent, methyl jasmonic acid reagent, crotonin reagent, and modified jasmonic acid derivatives.

[0094] The humidification device is a type of enclosed device equipped with an inlet / outlet air pump, a temperature and humidity sensor, a humidification device, and a temperature control device. The humidification device is used to regulate the ventilation state inside the device by adjusting the inlet / outlet air pump, monitor the temperature and humidity state inside the device by the temperature and humidity sensor, regulate the humidity state inside the device by the humidification device, and regulate the temperature state inside the device by the temperature control device.

[0095] The freezing device is a type of ultra-low temperature freezing equipment, with a freezing temperature that can reach -80℃;

[0096] The list of observed substances includes multiple categories of observed substances; each category of observed substances is an upstream or downstream metabolite in the synthetic pathway of the target substance.

[0097] Preferably, the step of sampling the experimental material once every sampling time interval from the start observation time to obtain a corresponding sample slice set based on a preset frozen section sampling rule specifically includes:

[0098] During the first sampling, the experimental material is removed from the humidification device; the tail end or head end of the experimental material is selected as the initial starting position; pruning shears are used to cut off and discard branches, roots, or stems approximately 0.5 cm after the initial starting position, and the current cut position of the experimental material is taken as the starting position of the current sampling; pruning shears are used again to cut off branches, roots, or stems approximately 0.5 cm after the initial sampling position, and the cut branches, roots, or stems are taken as the sampling material for this sampling; the previous cut position is marked on the current cut position of the experimental material; the marked experimental material is returned to the humidification device; and cryosectioning technology is used to continuously slice the sampling material according to a preset single slice thickness to obtain multiple corresponding sampling slices to form the corresponding sampling slice set; the thickness of each single slice is not less than 10 μm.

[0099] In each subsequent sampling, the experimental material is removed from the humidification device; the previously marked cutting position is taken as the starting position for the current sampling; pruning shears are used to cut and discard branches, roots, or stems approximately 0.5 cm after the current starting position, and the current cutting position of the experimental material is taken as the starting position for the current sampling; pruning shears are used again to cut branches, roots, or stems approximately 0.5 cm after the current sampling starting position, and the cut branches, roots, or stems are taken as the current sampling material; the previous cutting position is marked on the current cutting position of the experimental material; the marked experimental material is returned to the humidification device; and cryosectioning technology is used to continuously slice the current sampling material according to the thickness of a single slice to obtain multiple corresponding sampling slices to form the corresponding sampling slice set.

[0100] Preferably, the step of extracting the corresponding sampling data sequence from the sampling data of various types of observed substances in the observed substance list based on all the first slice subsets specifically includes:

[0101] Step 91: Set a preset mass-to-charge ratio sequence for the mass spectrometry imaging device based on the observed substance list; and use the mass spectrometry imaging device to perform spatial mass spectrometry imaging on each sampling slice of each first slice subset based on the mass-to-charge ratio sequence to obtain the corresponding first mass spectrum; and perform image alignment on all the first mass spectra, and spatially align all the image-aligned first mass spectra with the d-dimensional space of the first equation.

[0102] The mass spectrometry imaging device is a analytical ionization mass spectrometry imaging device.

[0103] The total number of observed substances in the observed substance list is denoted as A; each type of observed substance corresponds to a mass-to-charge ratio m / z.a 1 ≤ index a ≤ A; the total number of mass-to-charge ratios in the mass-to-charge ratio sequence is denoted as B, consisting of multiple mass-to-charge ratios m / z. b The order is determined by sequential sorting, where 1 ≤ index b ≤ B, and the mass-to-charge ratio m / z is between every two adjacent elements. b The difference between them is a preset first mass-to-charge ratio interval Δm / z1; the mass-to-charge ratio sequence covers all the mass-to-charge ratios m / z. a ;

[0104] All the first mass spectra have the same image size; each first mass spectra consists of multiple first pixels; each first pixel includes pixel coordinates, the mass-to-charge ratio sequence, and a signal intensity sequence S; the pixel coordinates are two-dimensional or three-dimensional coordinates; the signal intensity sequence S consists of B signal intensities s. b The signal strength s is ordered sequentially. b With the mass-to-charge ratio m / z b One-to-one correspondence;

[0105] After image alignment, the pixel coordinates of the two first pixels corresponding to the same slice point on any two first mass spectra are the same, wherein the same slice point refers to the point on each slice that has the same relative position with respect to the slice's central axis; after spatial alignment, the two first pixels corresponding to the same slice point on any two first mass spectra correspond to the same spatial point in the d-dimensional space.

[0106] Step 92: Take each of the first slice subsets as the corresponding current subset; and calculate the mean signal intensity of all the first mass spectra in the current subset to obtain the corresponding mean mass spectra;

[0107] The total number of mass spectra of the current subset is denoted as P;

[0108] The total number of pixels in each of the first mass spectra is equal, denoted as the total number Q;

[0109] Each of the first mass spectra consists of Q first pixels; the signal intensity sequence of each first pixel in each first mass spectra is denoted as S. p,q 1 ≤ index p ≤ P, 1 ≤ index q ≤ Q; the signal strength sequence S p,q Including B signal strengths

[0110] The mean mass spectrum is also composed of Q first pixels; the first pixels of the mean mass spectrum correspond one-to-one with the first pixels of each first mass spectrum corresponding to the current subset; the signal intensity sequence of each first pixel of the mean mass spectrum is denoted as S. q Each of the signal strength sequences Sq For the corresponding P signal intensity sequences S p,q The mean sequence; the signal strength sequence S q Including B signal strengths

[0111] The mean mass spectrum corresponds one-to-one with the first slice subset, and the total number of the mean mass spectra is equal to the total number of sampling points N. C ;

[0112] Step 93: Take each type of observed substance in the observed substance list as the corresponding current substance; and based on N... C The mean mass spectrum is the sampling data sequence corresponding to the current substance.

[0113] Preferably, the N-based C The mean mass spectrum is the sampling data sequence corresponding to the current substance, specifically including:

[0114] Step 101, set the mass-to-charge ratio m / z corresponding to the current substance. a As the current mass-to-charge ratio m / z * And based on the current mass-to-charge ratio m / z * Set the current mass-to-charge ratio range;

[0115] Wherein, the current mass-to-charge ratio range is [m / z * -△m / z2,m / z * +△m / z2];△m / z2 is the preset second mass-to-charge ratio interval, △m / z2>0;

[0116] Step 102: Use each of the mean mass spectra as the corresponding current mass spectra;

[0117] Step 103, convert each of the signal intensity sequences S of the current mass spectrum. q The sequence segment corresponding to the current mass-to-charge ratio interval is extracted as the current segment; and the maximum signal intensity in the current segment is selected. As the corresponding peak intensity For the current segment excluding the peak intensity The mean is calculated by taking the mean and standard deviation of all signal strengths except those mentioned above. and standard deviation And based on the peak intensity The mean and the standard deviation Calculate the signal-to-noise ratio S / N a , Regarding the signal-to-noise ratio S / N aDoes it exceed the preset signal-to-noise ratio threshold S / N? hold If the peak intensity is detected, then the peak intensity will be... As the corresponding signal strength Otherwise, set the corresponding signal strength. =0; and determined by the pixel coordinates corresponding to each of the first pixels in the current mass spectrum and the signal intensity. A corresponding second pixel is formed; and the Q second pixels are combined to form a signal intensity map corresponding to the current substance.

[0118] The signal strength map is composed of Q second pixels; each second pixel includes the pixel coordinates and the signal strength.

[0119] Step 104: For all non-zero signal strengths on the signal strength graph... The mean μ is obtained by calculating the mean and standard deviation. a and standard deviation σ a ; and based on the mean μ a The standard deviation σ a and the signal-to-noise ratio threshold S / N hold Calculate the corresponding signal strength threshold The signal strength graph is greater than zero but less than the signal strength threshold. The signal strength Reset to 0;

[0120] Step 105: Take each of the second pixels in the signal strength map as the corresponding current pixel; and determine the neighborhood space of the current pixel with the current pixel as the center and a preset neighborhood radius; and check all the signal strengths in the current neighborhood space. Perform mean calculation and use the result as the abundance corresponding to the current pixel. And for all the abundances obtained Normalization calculations are performed to obtain the corresponding normalized abundance. And the pixel coordinates corresponding to each of the second pixels in the signal intensity map and the normalized abundance. A corresponding third pixel is formed; and the Q third pixels are combined to form a normalized abundance map corresponding to the current substance.

[0121] The normalized abundance map is composed of Q third pixels; each third pixel includes the pixel coordinates and the normalized abundance.

[0122] Step 106: The non-zero normalized abundance values ​​on the normalized abundance map are... Extract them to form the corresponding first set; and use the total abundance of the first set as the corresponding total number of samples N. i ; and the normalized abundance of each of the first set As a corresponding substance abundance and the normalized abundance of each of the first set The spatial coordinate x of the corresponding spatial point of the third pixel in the d-dimensional space is used as the corresponding material coordinate. And by the coordinates of each of the aforementioned substances and the corresponding abundance of the substances To form a corresponding sample data The current mass spectrum is used as the corresponding sampling time t within the continuous time period [0, T]. i ; and the sampling time t obtained this time i and N i The sample data Form a corresponding sample set C of the sampling points i ;

[0123] Step 107, by N C The sample set C corresponding to the mean mass spectrum. i The samples are sequentially arranged to form a sequence of sampling data corresponding to the current substance.

[0124] Preferably, the step of performing targeted training on the model parameters of the two models based on the sampling data sequences of various observed substances and the model training process to obtain the corresponding first version of the parameter set specifically includes:

[0125] Step 111: Take each type of observed substance as the corresponding current substance; and take the sampling data sequence corresponding to the current substance as the current sequence;

[0126] Step 112, for the total number of samples N of the current sequence C Confirm; and extract the corresponding N from the current sequence. C The sample set of the sampling points And initialize a corresponding continuous convergence counter for the current sequence to 0; and based on the initialized model parameters The model parameters θ of the first model and the second model 1 θ 2 Perform initialization;

[0127] Step 113: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T The discrete time points t n ; and the start time t of the current sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 At each spatial point location in the d-dimensional space, N0 virtual particles p are initialized. k ; and for each of the virtual particles p k Initialize a set of corresponding particle motion trajectories and the mass change trajectory And each of the virtual particles p k The coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and apply the mass weight initialization rule to each virtual particle p. k Quality weight Perform initialization;

[0128] Step 114: Substitute the first model inference expression into the second equation, solve the second equation using the Euler method, and complete the determination of the motion trajectory of each particle by solving the equation. The effective coordinates are filled; the second model inference expression is substituted into the third equation, and the third equation is solved using the Euler method. The mass change trajectories are then solved by solving the equations. Effective weight filling;

[0129] Step 115, obtain the particle motion trajectories of group N0. and the mass change trajectory and N C The sample set of the sampling points Substitute the loss function L ALL Obtain the corresponding current loss value;

[0130] Step 116: Identify whether the current loss value meets the convergence loss value range; if not, clear the continuous convergence counter and proceed to step 117; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the continuous convergence count threshold; if it exceeds, proceed to step 118; if it does not exceed, proceed to step 117.

[0131] Step 117, based on the model optimizer, move towards making the first loss function, the second loss function, the third loss function, and the loss function L... ALL The direction in which both reach their minimum values ​​is related to the model parameters θ of the first and second models.1 θ 2 Perform one round of modulation; and return to step 114 when the current round of modulation is completed;

[0132] Step 118: Stop the current training and use the current model parameters θ of the first model and the second model. 1 θ 2 Form a first version parameter set corresponding to the current substance.

[0133] Preferably, the simulation of the spatiotemporal change state of the current matter based on the first set of parameters for various types of observed matter specifically includes:

[0134] Step 121: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T +1 of the discrete time points t n 0 ≤ index n ≤ N T ; and based on N T +1 of the discrete time points t n Construct a spacetime system H with the d-dimensional space;

[0135] The d-dimensional space includes N0 spatial points, and the coordinates of each spatial point are denoted as x. o , 1 ≤ index o ≤ N0; the spatiotemporal system H is composed of (N T +1)×N0 spacetime points Composition; each of the aforementioned spatiotemporal points The spatiotemporal characteristics include the corresponding discrete time point t n and the coordinate x o ;

[0136] Step 122: Take each type of observed substance in the observed substance list as the corresponding current substance;

[0137] Step 123: Take the first version of the parameter set corresponding to the current substance as the corresponding current parameter set; and adjust the model parameters θ of the first model and the second model based on the current parameter set. 1 θ 2 Perform the settings; and use the first model and the second model obtained from this setting as the corresponding current first model and current second model;

[0138] Step 124, from each of the aforementioned spacetime points The discrete time point t n and the coordinate x oAs a set of corresponding first times and first coordinates, a corresponding first spatiotemporal vector is input into the current first model for processing, and the first predicted velocity field obtained from this processing is used as a corresponding velocity field. And by each of the aforementioned spacetime points The discrete time point t n and the coordinate x o As a set of corresponding second times and second coordinates, a corresponding second spatiotemporal vector is input into the current second model for processing, and the first predicted net generation rate obtained from this processing is used as a corresponding net generation rate.

[0139] Step 125, (N) T +1)×N0 of the aforementioned spacetime points The corresponding discrete time point t n The coordinate x o The velocity field and the net generation rate Substituting the first equation yields the corresponding current partial differential equation; and solving the current partial differential equation yields each of the spatiotemporal points. Corresponding material density

[0140] The current partial differential equation is:

[0141]

[0142] Step 126, from each of the aforementioned spacetime points The corresponding discrete time point t n The coordinate x o The net production rate and the density of the substance This forms a corresponding spatiotemporal feature vector; and is derived from the (N) obtained in this study. T +1)×N0 of the spatiotemporal point feature vectors form a first simulation data corresponding to the current substance.

[0143] Preferably, the step of identifying the synthetic central region and non-central comparison regions of the second slice subset of various observed materials based on simulated data specifically includes:

[0144] Step 131: Take each type of observed substance in the observed substance list as the corresponding current substance;

[0145] Step 132: Take the first simulation data corresponding to the current substance as the current simulation data; and extract the corresponding (N) from the current simulation data. T+1)×N0 spatiotemporal point feature vectors; and extract the corresponding discrete time point t from each of the spatiotemporal point feature vectors. n The coordinate x o The net production rate and the density of the substance

[0146] Step 133, based on the obtained (N) T +1)×N0 group's net generation rate and the density of the substance The time point of strongest activity of the current substance Perform identification;

[0147] Among them, the strongest activity time point and the corresponding index n * The identification method is as follows:

[0148]

[0149] Step 134, the discrete time point t in the current simulation data n With the strongest activity time point The matched spatiotemporal point feature vectors are extracted to form a corresponding second set; and the metabolic flux corresponding to each spatiotemporal point feature vector in the second set is... Perform calculations; and for the obtained N0 metabolic fluxes... The mean μ is obtained by calculating the mean and standard deviation. * and standard deviation σ * ; and based on the mean μ * The standard deviation σ * Calculate the corresponding metabolic flux threshold γ hold ;

[0150] Wherein, the metabolic flux The calculation method is as follows:

[0151]

[0152] The metabolic flux threshold γ hold The calculation method is as follows:

[0153] γ hold =μ * +β·σ * ,

[0154] The standard deviation factor β is a preset positive integer;

[0155] Step 135, the metabolic flux threshold γ is greater than or equal to the threshold value. hold The metabolic flux Let γ be the corresponding high-activity flux. high ; and each of the aforementioned high-activity flux γ high The corresponding coordinate x o Let x be the coordinates of the corresponding high-activity point. high ; and for all the aforementioned highly active point coordinates x high All corresponding spatial points are clustered into multiple corresponding first point clusters; the total number of spatial points in each first point cluster is counted to obtain the corresponding first total number; and the first point clusters whose first total number is greater than or equal to a preset threshold for the total number of points are recorded as the corresponding second point clusters.

[0156] Step 136, for all the highly active fluxes γ in each of the second point clusters high The corresponding flux sum is obtained by summing the flux sums; the second point cluster corresponding to the largest flux sum is taken as the corresponding center point cluster; and region edge identification is performed in the d-dimensional space based on the center point cluster, and the spatial region of the currently identified region edge is taken as the current material at the time point of strongest activity. The active center region on;

[0157] Step 137: Combine the sampling times of all the second slices corresponding to the current substance with the time point of strongest activity. The closest second slice subset is taken as the most active subset; and a high-resolution image of each sample slice of the most active subset is obtained by using a preset optical microscope, which is recorded as the first slice image; and all the first slice images are image aligned, and all the image-aligned first slice images are spatially aligned with the d-dimensional space.

[0158] The optical microscope is a high-resolution optical microscope equipped with optical slicing technology.

[0159] All the first slice images have the same image size; each first slice image consists of multiple fourth pixels;

[0160] After image alignment, the pixel coordinates of the two fourth pixels corresponding to the same slice point on any two first slice images are the same; after spatial alignment, the two fourth pixels corresponding to the same slice point on any two first slice images correspond to the same spatial point in the d-dimensional space.

[0161] Step 138: The image region corresponding to the active center region on each of the first slice images is recorded as the corresponding synthesis center region; and based on a preset comparison region selection rule, an image region outside the synthesis center region and with the same shape as the synthesis center region is selected on each of the first slice images as the corresponding non-center comparison region; and each of the first slice images marked with the synthesis center region and the non-center comparison region is recorded as the corresponding second slice image; and all the second slice images corresponding to the current substance are combined to form a corresponding region identification data.

[0162] The alignment region selection rule by default requires that the cell types contained in the non-center alignment region be consistent with those in the synthesis center region, and requires that the non-center alignment region does not contain any of the high-activity point coordinates x in the corresponding spatial region in the d-dimensional space. high Furthermore, the shape of the non-center alignment region must be consistent with that of the synthesis center region.

[0163] Preferably, the step of identifying the corresponding key gene set by comparing the gene sequencing results of the synthesis centers and comparison regions of various observed substances specifically includes:

[0164] Step 141: Take each type of observed substance in the observed substance list as the corresponding current substance;

[0165] Step 142: Use the region identification data of the current substance as the current identification data; extract multiple second slice images marked with the synthesis center region and the non-center comparison region from the current identification data; and use the strongest active subset corresponding to the current substance as the corresponding current subset;

[0166] The second slice image corresponds one-to-one with the sampled slice of the current subset;

[0167] Step 143: Take each of the second slice images as the corresponding current slice image; use a preset laser cutting microscope to align the edges of the current slice image and the corresponding sampling slice with the object; and after alignment, cut the current sampling slice into corresponding regional samples according to the synthetic center region and the non-center comparison region marked on the current slice image to obtain a set of corresponding synthetic center region samples and non-center comparison region samples to form a corresponding first sample group.

[0168] Step 144: Based on high-throughput sequencing technology, transcriptome sequencing is performed on the synthetic central region samples and the non-central alignment region samples of each of the first sample groups to obtain corresponding central region sequencing data and alignment region sequencing data; differential gene expression analysis is performed on the central region sequencing data and the alignment region sequencing data to obtain a set of differentially expressed genes with an upregulated gene reading in the central region sequencing data; GO enrichment analysis is performed on the differentially expressed gene set to screen the transcriptional regulatory factors of the current substance to obtain a corresponding set of first transcriptional regulatory factors; and KEGG pathway enrichment is performed on the differentially expressed gene set to screen the metabolic enzyme genes of the synthetic transformation pathway of the current substance to obtain a corresponding set of first metabolic enzyme genes.

[0169] Step 145: Take the intersection of the first transcription regulatory factor sets of all the first sample groups to obtain the corresponding second transcription regulatory factor sets; and take the intersection of the first metabolic enzyme gene sets of all the first sample groups to obtain the corresponding second metabolic enzyme gene sets; and the second transcription regulatory factor sets and the second metabolic enzyme gene sets constitute the corresponding key gene sets.

[0170] A second aspect of the present invention provides an apparatus for implementing the analysis method for plant metabolite synthesis centers and key genes described in the first aspect above. The apparatus includes: an analysis tool preparation module, an experimental material preparation module, an observation and sampling module, an adaptive training module, a simulation analysis module, and a report generation module.

[0171] The analysis tool preparation module is used to construct partial differential equations to describe how the state of matter changes based on the spatiotemporal changes of the state of matter transportation and the state of matter synthesis or consumption, and to denote the current partial differential equation as the first equation; to design corresponding neural network models for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation, and to denote the corresponding first model and second model; to set two corresponding ordinary differential equations for the first equation, and to denote the corresponding second equation and third equation; and to customize the data format for the sampled data sequence loaded with experimental sampling data, and to design the corresponding loss function L. ALL Based on the sampled data sequence, the second equation, the third-party process, and the loss function L... ALL Design a model training process for training the first model and the second model;

[0172] The experimental material preparation module is used to extract the synthesis site of the target substance from the plant material as the experimental material; and to select an experimental reagent to spray the experimental material; and to store the treated experimental material in a preset moisturizing device.

[0173] The observation and sampling module is used to take the material preservation start time as the start observation time, and customize the sampling time axis based on the preset observation duration and sampling time interval; and from the start observation time, every sampling time interval, the experimental material is sampled once based on the preset frozen section sampling rules to obtain the corresponding sampled section set; and the current sampled section set is divided into a first section subset and a second section subset and placed in a preset freezing device for storage;

[0174] The adaptive training module is used to construct an observation substance list based on all upstream and downstream metabolites along the target substance synthesis pathway after sampling; extract sampling data from each type of observation substance in the observation substance list based on all the first slice subsets to obtain the corresponding sampling data sequence; and perform targeted training on the model parameters of the two models based on the sampling data sequence of each type of observation substance and the model training process to obtain the corresponding first version of the parameter set.

[0175] The simulation analysis module simulates the spatiotemporal change state of the current material based on the first set of parameters for each type of observed material; and identifies the synthetic center region and non-center comparison region of the second slice subset of each type of observed material based on the simulation data; and identifies the corresponding key gene set by comparing the gene sequencing results of the synthetic center and the comparison region of each type of observed material.

[0176] The report generation module constructs corresponding substance analysis reports based on the synthesis centers and key gene analysis information of various observed substances; and the metabolite analysis report corresponding to the target substance is composed of all the substance analysis reports and saved.

[0177] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0178] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0179] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0180] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0181] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for analyzing plant metabolite synthesis centers and key genes. As described above, this invention pre-constructs a partial differential equation (i.e., the first equation) to describe how the state of matter changes based on the spatiotemporal variations of the state of matter transport and the state of matter synthesis or consumption. Two artificial intelligence models (i.e., the first model and the second model) are designed for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation. Driven by the velocity field spatiotemporal function and the net generation rate spatiotemporal function of the first equation, two ordinary differential equations (the second equation and the third equation) are set to simulate the coordinates and mass weights of virtual particles from a Lagrange perspective. The solution process of the two ordinary differential equations is combined to design the corresponding loss function L for the joint training of the first and second models. ALL Based on the sampled data sequence, two ordinary differential equations, and the loss function L... ALL A standard processing flow for joint training (i.e., model training flow) is designed. This embodiment of the invention obtains multiple sampling slice sets at various observation times through experimental observation, and divides each sampling slice set into two subsets (first slice subset and second slice subset). The first slice subset at all observation times is used for the analysis of the strongest activity time point and the location of the synthesis center, while the second slice subset closest to the strongest activity time point is used for key gene information analysis. In analyzing the strongest activity time point and the location of the synthesis center, this embodiment first constructs a list of observed substances based on all known upstream and downstream metabolites along the target substance's synthesis pathway. Then, based on all first slice subsets, the sampling data sequences of various observed substances are constructed. Next, based on the sampling data sequences of various observed substances and the standard model training flow, the model parameters of the two models are trained specifically to obtain a set of model parameters corresponding to the current substance (i.e., the first set of parameters). Finally, based on the model parameter sets of various observed substances, spatiotemporal change state simulations are performed, and the strongest activity time point and the location of the synthesis center for each observed substance are confirmed by analyzing the spatiotemporal simulation data. In analyzing key gene information of various observed substances, this invention selects the second subset of slices closest to the sampling time and the time point of strongest activity as the analysis subset for the current substance. Central / alignment region samples are extracted from the sampled slices in the current analysis subset using optical and laser cutting microscopes. Transcriptome sequencing is then performed on the central / alignment region samples, and the corresponding key gene information is identified through alignment sequencing results. This invention achieves the technical objective of locating the time point of strongest activity and the center of synthesis for each type of metabolite in a spatiotemporal system. The adaptive model training process corresponding to each type of observed substance improves model generalization and localization accuracy. Furthermore, this invention improves the precision and accuracy of central / alignment region sample extraction, thereby enhancing the accuracy of key gene information identification. Attached Figure Description

[0182] Figure 1 This is a schematic diagram of an analysis method for plant metabolite synthesis centers and key genes provided in Embodiment 1 of the present invention;

[0183] Figure 2 This is a modular structure diagram of an analysis device for plant metabolite synthesis centers and key genes provided in Embodiment 2 of the present invention;

[0184] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0185] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0186] Embodiment 1 of the present invention provides a method for analyzing the biosynthetic centers and key genes of plant metabolites, such as... Figure 1 The schematic diagram shows a method for analyzing plant metabolite synthesis centers and key genes provided in Embodiment 1 of the present invention. The method mainly includes the following steps:

[0187] Step 1: Construct partial differential equations to describe how the state of matter changes based on the spatiotemporal changes of the state of matter transportation and the state of matter synthesis or consumption, and denote the current partial differential equation as the first equation; design corresponding neural network models for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation, and denote them as the first model and the second model; set two corresponding ordinary differential equations for the first equation, and denote them as the second equation and the third equation; customize the data format for the sampling data sequence loaded with experimental sampling data; and design the corresponding loss function L. ALL Based on the sampled data sequence, the second equation, the third process, and the loss function L... ALL Design a model training process for training the first and second models;

[0188] Specifically, it includes: Step 11, constructing a partial differential equation to describe how the state of matter changes based on the spatiotemporal changes of the state of matter transportation in space and the state of matter synthesis or consumption, and denoting the current partial differential equation as the first equation.

[0189] Here, the first equation of this embodiment of the invention is:

[0190]

[0191] Where t is time and x is the coordinate of a point in space;

[0192] Time t is a point in time within a preset continuous time period [0,T], and T is a preset time length; here, time length T = observation duration - sampling time interval. The observation duration and sampling time interval are two preset time length parameters that will be used in the experimental observation process below. The observation duration is an integer multiple of the sampling time interval.

[0193] The coordinate x of a spatial point is the coordinate of a point in a predefined d-dimensional space, where d is 2 or 3; the d-dimensional space is a network of spatial points, consisting of multiple evenly distributed spatial points, each corresponding to a d-dimensional coordinate; the total number of spatial points in the d-dimensional space is denoted as N0.

[0194] ρ(t,x) is a density spacetime function with the dimension of matter / space, used to describe the density of the corresponding spatial point at time t at the coordinate x.

[0195] v(t,x) is the spacetime function of the velocity field, with the dimension of space / time. It is used to describe the d-dimensional velocity field of the corresponding spatial point at time t. The d-dimensional velocity field consists of d-dimensional material outflow velocity components. If each component is greater than 0, it indicates a net outflow of material in the current dimension; if it is less than 0, it indicates a net inflow of material in the current dimension.

[0196] g(t,x) is a spacetime function of net generation rate with the dimension 1 / time. It is used to describe the net generation rate of the corresponding spatial point at time t, which represents the proportion of change in the amount of matter caused by the synthesis or consumption of local matter at the current spatial point per unit time.

[0197] This is the local time derivative term of the partial differential equation, used to describe the rate of change of density of the corresponding spatial point at time t at coordinate x.

[0198] Here, v(t,x)·ρ(t,x) is the divergence term of the partial differential equation, used to describe the mass flux at time t at the spatial point corresponding to x. Used to describe the spatial divergence of mass flux; a spatial divergence greater than 0 indicates a net outflow of mass, and a spatial divergence less than 0 indicates a net inflow of mass.

[0199] g(t,x)·ρ(t,x) are the source and sink terms of the partial differential equation, where g(t,x) is the source term and ρ(t,x) is the sink term. The source and sink terms are used to describe the density change of the spatial point corresponding to x at time t caused by the synthesis or consumption of local matter at the spatial point, without considering the factors of matter flow.

[0200] Step 12, and design corresponding neural network models for the velocity field spacetime function and the net generation rate spacetime function in the first equation, denoted as the first model and the second model respectively;

[0201] Here, the first model and the second model in this embodiment of the invention are each implemented based on a type of neural network model used for processing regression calculation tasks; the neural network model used for processing regression calculation tasks includes MLP models and CNN models;

[0202] The first model in this embodiment of the invention is used to predict the velocity field based on the first spatiotemporal vector input to the model and output the corresponding first predicted velocity field; wherein, the first spatiotemporal vector includes a first time and a first coordinate; the first time and the first coordinate correspond to the time t and spatial point coordinate x of the velocity field spatiotemporal function v(t,x), respectively; the first predicted velocity field includes d first predicted velocity components; the model inference expression of the first model is: v * =F 1 (t,x;θ 1 );v * For the first predicted velocity field, F 1 () represents the reasoning process of the first model, θ 1 These are the model parameters for the first model;

[0203] The second model in this embodiment of the invention is used to predict the net generation rate based on the second spatiotemporal vector input to the model and output the corresponding first predicted net generation rate; wherein, the second spatiotemporal vector includes a second time and a second coordinate; the second time and the second coordinate correspond to the time t and spatial point coordinate x of the spatiotemporal function g(t,x) of the net generation rate, respectively; the model inference expression of the second model is: g * =F 2 (t,x;θ 2 ); where g * F is the first prediction ratio to the net generation rate. 2 () represents the reasoning process of the second model, θ 2 These are the model parameters for the second model;

[0204] Step 13, and set up two corresponding ordinary differential equations for the first equation, denoted as the corresponding second equation and third equation;

[0205] Specifically, this includes: Step 131, constructing N0 virtual particles p in d-dimensional space based on the Lagrange perspective.k 1 ≤ index k ≤ N0; and for each virtual particle p k Set the corresponding coordinates and quality weight And each virtual particle p k Coordinates at the start of the continuous time interval [0, T] Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and initialize each virtual particle p according to the preset mass weight initialization rule. k Mass weight at the start of the continuous time interval [0,T] Perform initialization;

[0206] Among them, virtual particle p k The starting time of the continuous time interval [0,T] corresponds one-to-one with a point in the d-dimensional space; coordinates Let be the real-time coordinates of the k-th virtual particle in the d-dimensional space at time t, and let be the mass weight. Let be the total amount of matter carried by the k-th virtual particle at time t;

[0207] Step 132: Design two ordinary differential equations for each virtual particle p based on the first equation. k The coordinates at any time t within the continuous time interval [0,T] and quality weight Evolution;

[0208] Among them, the two ordinary differential equations include the second equation and the third equation;

[0209] The second equation is a flow mapping equation for the time-varying coordinates of a virtual particle, driven by the velocity field spacetime function v(t,x), specifically:

[0210]

[0211] The solution to the second equation is For any virtual particle p, the time evolution mapping is used. k In terms of time evolution mapping Corresponding to a dataset This dataset contains the particle trajectory of the current virtual particle.

[0212] Substituting the first model inference expression corresponding to the velocity field spacetime function v(t,x) into the second equation, we get:

[0213]

[0214] The third equation is a flow mapping equation for the evolution of the mass weight of virtual particles over time, driven by the spacetime function g(t,x) of the net generation rate. Specifically:

[0215]

[0216] The solution to the third equation is For any virtual particle p, the time evolution mapping of the mass weights is used. k In terms of time evolution mapping Corresponding to a dataset This dataset represents the trajectory of the current virtual particle's mass change.

[0217] Substituting the second model inference expression corresponding to the net generation rate spatiotemporal function g(t,x) into the third process, we obtain:

[0218]

[0219] There are various methods for solving the second and third equations, such as the Euler method, the improved Euler method, and the fourth-order Runge-Kutta method; the embodiment of this invention provides the solution process of the Euler method, as shown below:

[0220] 1) The process of solving the second equation using Euler's method is as follows:

[0221] The continuous time period [0, T] is divided into multiple discrete time points t according to a preset time interval Δt. n 0 ≤ index n ≤ N T The total number of discrete points N T = T / △t; where the time interval △t is a preset time length parameter in this embodiment of the invention, and the time interval △t < sampling time interval;

[0222] And the second equation Approximately:

[0223]

[0224] When n=0,

[0225] Virtual particle p k coordinates Given a quantity, then:

[0226]

[0227] When n=1,

[0228]

[0229] By iterating point by point in this manner, the virtual particle p is obtained. k N in the continuous time interval [0,T] T +1 coordinates Forming the corresponding particle motion trajectory

[0230] 2) In the trajectory of particle motion Given the given information, the process of solving the third equation using the Euler method is as follows:

[0231] Third-party process Approximately:

[0232]

[0233] When n=0,

[0234] Virtual particle p k Quality weight Given a quantity, then:

[0235]

[0236] When n=1,

[0237]

[0238] By iterating point by point in this manner, the virtual particle p is obtained. k N in the continuous time interval [0,T] T +1 quality weight The corresponding mass change trajectory

[0239] Step 14, and customize the data format for the sampled data sequence for loading experimental sampled data;

[0240] Here, the data format of the sampling data sequence in this embodiment of the invention is as follows:

[0241] The sampled data sequence consists of a sample set C of multiple sampling points. i Composition, 0 ≤ index i ≤ N C N C The total number of sampling points; the sample set C for each sampling point. i Including a sampling time t i and multiple sample data Sampling time t i Let j be a moment in the continuous time interval [0, T], where 1 ≤ index j ≤ N. i N i For the sample set C of the sampling points i Total number of samples; sample data Including material coordinates Material abundance

[0242] Step 15, and design the corresponding loss function L. ALL ;

[0243] Here, the loss function L in the embodiment of the invention ALLfor:

[0244] L ALL =λ1L1+λ2L2+λ3L3;

[0245] Where λ1, λ2, and λ3 are the preset first, second, and third loss weights, and L1, L2, and L3 are the corresponding first, second, and third loss functions;

[0246] 1) The first loss function L1 is a regularization loss, specifically:

[0247]

[0248] Where, λ v , λ g These are two learnable weight parameters;

[0249] 2) The second loss function L2 is a quality loss, specifically:

[0250]

[0251]

[0252] in, For all virtual particles at sampling time t i The predicted total amount of matter above, The sampling time t is calculated based on the sampled data sequence. i The total amount of matter on the surface For the first sampling time t i＝1 Corresponding observations; f m→w () is the preset abundance-quality weighting conversion function;

[0253] Quality weight The calculation method is as follows:

[0254] For the known mass change trajectory Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n Corresponding quality weights As the current quality weight If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let time t be denoted as t - t + , t - <t i <t +And based on the time t before and after - t + The two corresponding quality weights The current quality weight is obtained by performing linear interpolation.

[0255] 3) The third loss function L3 is an optimal transmission loss, specifically:

[0256]

[0257]

[0258] Among them, F dis () represents the preset Wasserstein distance function; The total mass of all virtual particles at sampling time t i Discrete distribution predictor on, The sampling time t is calculated based on the sampled data sequence. i Discrete distribution observations on; δ is the Dirac measure, Dirac measure coordinates The unit point mass of the corresponding spatial point, Dirac measure coordinates The unit point mass of the corresponding spatial point;

[0259] coordinate The calculation method is as follows:

[0260] For known particle trajectories Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n Corresponding coordinates As current coordinates If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let time t be denoted as t - t + , t - <t i <t + And based on the time t before and after - t + The two corresponding coordinates Perform d-dimensional coordinate interpolation calculations and use the results as the current coordinates.

[0261] Step 16, based on the sampled data sequence, the second equation, the third equation, and the loss function L. ALL Design a model training process for training the first and second models;

[0262] Specifically, this includes: Step 161, calculating the total number N samples in the current sampled data sequence. C Confirm; and extract the corresponding N from the sequence. C Sample set of sampling points The continuous convergence counter is initialized to 0; and a set of preset initialization model parameters are used. For the model parameters θ of the first and second models 1 θ 2 Perform initialization;

[0263] Step 162: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T Discrete time point t n ; and the start time t of the sampled data sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 Initialize N0 virtual particles p at each spatial point in the d-dimensional space. k ; and for each virtual particle p k Initialize a set of corresponding particle motion trajectories and mass change trajectory And each virtual particle p k coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and initialize each virtual particle p according to the mass weight initialization rule. k Quality weight Perform initialization;

[0264] Here, the quality weight initialization rule in this embodiment of the invention is:

[0265] All quality weights Use the current weight as the coordinates corresponding to the current weight. Used as the current coordinates; and for the sample set of sampling points Perform a query; if a material coordinate exists... If the coordinates are the same as the current coordinates, then the abundance-quality weighting transformation function f is used. m→w () for the current material coordinates Corresponding material abundance Perform quality weight transformation and set the current weights based on the transformation results. If there are no material coordinates with the same current coordinates Then based on the preset default quality weight wdefault Set the current weights. Among them, the default quality weight w default It is a pre-set value;

[0266] It should be noted that the trajectories of each particle after initialization are... Except for the first coordinate All coordinates other than the valid coordinates are invalid coordinates; the mass change trajectories after initialization are complete. Except for the first quality weight All weights other than the valid weights are invalid weights.

[0267] Step 163: Substitute the reasoning expression of the first model into the second equation, solve the second equation using the Euler method, and complete the determination of the trajectory of each particle by solving the equation. The effective coordinates are filled; the reasoning expression of the second model is substituted into the third equation, and the third equation is solved by Euler's method. The trajectory of each mass change is completed by solving the equation. Effective weight filling;

[0268] Step 164, obtain the particle trajectories of group N0. and mass change trajectory and N C Sample set of sampling points Substitute the loss function L ALL Obtain the corresponding current loss value;

[0269] Step 165: Identify whether the current loss value meets the preset convergence loss value range; if not, clear the continuous convergence counter to zero and proceed to step 166; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the preset continuous convergence count threshold; if it exceeds the threshold, proceed to step 167; if it does not exceed the threshold, proceed to step 166.

[0270] Here, the convergence loss value range is a preset numerical range, and the threshold for the number of consecutive convergences is a preset threshold parameter.

[0271] Step 166: Based on the preset model optimizer, optimize the first loss function, the second loss function, the third loss function, and the loss function L. ALL The directions that both reach their minimum values ​​are related to the model parameters θ of the first and second models. 1 θ 2 Perform one round of modulation; and return to step 163 when the current round of modulation is completed;

[0272] The model optimizers include the Adam optimizer and the SGD optimizer.

[0273] Step 167: Stop the current training and use the current model parameters θ of the first and second models. 1 θ 2 This forms a corresponding first version of the parameter set.

[0274] Step 2: Extract the synthesis site of the target substance from the plant material as the experimental material; select an experimental reagent to spray the experimental material; and store the treated experimental material in a pre-set humidification device.

[0275] Specifically, this includes: Step 21, extracting the synthesis site of the target substance from the plant material as experimental material;

[0276] Here, the target substance in this embodiment of the invention is a class of terpenoid compounds;

[0277] In this embodiment of the invention, the target substance is synthesized from plant branches, roots, or stems.

[0278] Step 22, and select an experimental reagent to spray the experimental materials;

[0279] Here, the experimental reagents in this embodiment of the invention are a class of reagents that can stimulate the metabolic activity of target substances, including jasmonic acid reagent, methyl jasmonic acid reagent, crotonin reagent, and modified jasmonic acid derivatives;

[0280] Step 23, and then place the treated experimental materials in a pre-set humidification device for storage;

[0281] Here, the humidification device in this embodiment of the invention is a type of enclosed device with an inlet / outlet air pump, a temperature and humidity sensor, a humidification device, and a temperature control device; the humidification device is used to regulate the ventilation state inside the device by adjusting the inlet / outlet air pump, monitor the temperature and humidity state inside the device by the temperature and humidity sensor, regulate the humidity state inside the device by the humidification device, and regulate the temperature state inside the device by the temperature control device.

[0282] Step 3: Take the material preservation start time as the start observation time, and customize the sampling time axis based on the preset observation duration and sampling time interval; and sample the experimental material once every sampling time interval from the start observation time according to the preset frozen section sampling rules to obtain the corresponding sampling section set; and divide the current sampling section set into the first section subset and the second section subset and store them in the preset freezing device.

[0283] Specifically, this includes: Step 31, taking the material preservation start time as the start observation time, and customizing the sampling time axis based on the preset observation duration and sampling time interval;

[0284] Here, in this embodiment of the invention, the starting zero time of the continuous time period [0,T] corresponds to the first sampling point on the sampling time axis, and the time interval between the first sampling point and the starting observation time is the sampling time interval; the ending time of the continuous time period [0,T] corresponds to the last sampling point on the sampling time axis, and the last sampling point is the ending time of the sampling time axis; the total number of sampling points N on the sampling time axis is... C = Observation duration / Sampling time interval;

[0285] Step 32: Starting from the initial observation time, the experimental material is sampled once every sampling time interval based on the preset frozen section sampling rules to obtain the corresponding sampled section set;

[0286] Specifically, the process includes: Step 321, during the first sampling, removing the experimental material from the humidification device; selecting the tail end or head end of the experimental material as the initial starting position; using pruning shears to cut and discard branches, roots, or stems approximately 0.5 cm after the initial starting position, and taking the current cut position of the experimental material as the starting position for this sampling; again using pruning shears to cut branches, roots, or stems approximately 0.5 cm after the sampling starting position, and taking the cut branches, roots, or stems as the sampling material for this sampling; marking the previous cut position on the current cut position of the experimental material; returning the marked experimental material to the humidification device; and using cryosectioning technology to continuously slice the sampling material according to a preset single slice thickness to obtain multiple corresponding sampling slices forming a corresponding sampling slice set;

[0287] Here, the thickness of a single slice in this embodiment of the invention is not less than 10 μm;

[0288] Step 322: For each subsequent sampling, remove the experimental material from the humidification device; use the previously marked cut position as the starting position for the current sampling; use pruning shears to cut and discard branches, roots, or stems approximately 0.5 cm after the current starting position, and use the current cut position as the starting position for the current sampling; again use pruning shears to cut branches, roots, or stems approximately 0.5 cm after the sampling starting position, and use the cut branches, roots, or stems as the sampling material for this sampling; mark the previous cut position on the current cut position of the experimental material; put the marked experimental material back into the humidification device; and use cryosectioning technology to continuously slice the sampling material according to the thickness of a single slice to obtain multiple corresponding sampling slices to form a corresponding sampling slice set;

[0289] Step 33, and divide the current sampled slice set into a first slice subset and a second slice subset and store them in a preset freezing device;

[0290] Here, the freezing device in this embodiment of the invention is a type of ultra-low temperature freezing equipment, whose freezing temperature can reach -80°C; when preserving the sampled slice set, the freezing temperature of the freezing device is set to -80°C.

[0291] Step 4: After sampling, construct an observation material list based on all upstream and downstream metabolites along the target substance synthesis pathway; extract sampling data for each type of observation material in the observation material list based on all first slice subsets to obtain the corresponding sampling data sequence; and perform targeted training on the model parameters of the two models based on the sampling data sequences of each type of observation material and the model training process to obtain the corresponding first version of the parameter set.

[0292] Specifically, this includes: Step 41, after sampling, constructing an observation material list based on all upstream and downstream metabolites along the target material's synthesis pathway;

[0293] Here, the observation substance list of this invention includes multiple categories of observation substances; each category of observation substance is a known upstream or downstream metabolite on the synthetic pathway of the target substance; the observation substance list is actually a substance list composed of all known upstream and downstream metabolites on the known synthetic pathway of the target substance.

[0294] Step 42, and extract the corresponding sampling data sequence by sampling data of various types of observation materials in the observation material list based on all first slice subsets;

[0295] Specifically, it includes: step 421, setting a preset mass-to-charge ratio sequence for the mass spectrometry imaging device based on the observed material list; using the mass spectrometry imaging device to perform spatial mass spectrometry imaging on each sampling slice of each first slice subset based on the mass-to-charge ratio sequence to obtain the corresponding first mass spectrum; and performing image alignment on all first mass spectra, and spatially aligning all image-aligned first mass spectra with the d-dimensional space of the first equation.

[0296] Here, the mass spectrometry imaging device in this embodiment of the invention is a analytical ionization mass spectrometry imaging device;

[0297] The total number of observed substances in the observation substance list of this invention is denoted as A; each type of observed substance corresponds to a mass-to-charge ratio m / z. a , 1 ≤ index a ≤ A;

[0298] The total number of mass-to-charge ratios in the mass-to-charge ratio sequence of this invention is denoted as B, consisting of multiple mass-to-charge ratios m / z. b It is ordered sequentially, with 1 ≤ index b ≤ B, and the mass-to-charge ratio of every two adjacent elements is m / z. b The difference between them is the preset first mass-to-charge ratio interval Δm / z1; the mass-to-charge ratio sequence covers all mass-to-charge ratios m / z. a ;

[0299] It should be noted that the observed material list in this embodiment of the invention corresponds to only A discrete mass-to-charge ratios m / z. a The mass-to-charge ratio sequence is a uniformly spaced data sequence where B > A, and there are A mass-to-charge ratios m / z in the sequence. b With A mass-to-charge ratio m / z a One-to-one correspondence; during mass spectrometry imaging, the feedback of each pixel is a signal intensity-mass-charge ratio curve, which can also be regarded as a signal intensity sequence; given A m / z values... a Given △m / z1, A m / z values ​​can be represented. a The maximum and minimum values ​​in the middle are used as the upper and lower boundaries, and the upper and lower boundaries are floated upward and downward according to two pre-set upward and downward floating percentages. The floated upper and lower boundaries form a mass-to-charge ratio range. The mass-to-charge ratio sequence of the present invention can be obtained by uniformly sampling this range based on Δm / z1.

[0300] All first mass spectra in this embodiment of the invention have the same image size; each first mass spectra consists of multiple first pixels; each first pixel includes pixel coordinates, a mass-to-charge ratio sequence, and a signal intensity sequence S; the pixel coordinates are two-dimensional or three-dimensional coordinates; the signal intensity sequence S consists of B signal intensities s. b Arranged in sequence, signal strength s b With mass-to-charge ratio m / z b One-to-one correspondence;

[0301] It should be noted that after image alignment, the pixel coordinates of the two first pixels corresponding to the same slice point on any two first mass spectra are the same. Here, the same slice point refers to the point on each slice that has the same relative position with respect to the slice's central axis. After spatial alignment, the two first pixels corresponding to the same slice point on any two first mass spectra correspond to the same spatial point in d-dimensional space.

[0302] Step 422: Take each first slice subset as the corresponding current subset; and calculate the mean signal intensity of all first mass spectra in the current subset to obtain the corresponding mean mass spectra.

[0303] Here, the total number of mass spectra of the current subset is denoted as P;

[0304] The total number of pixels in each of the first mass spectra is equal, denoted as the total number Q;

[0305] Each first mass spectrum consists of Q first pixels; the signal intensity sequence of each first pixel in each first mass spectrum is denoted as S. p,q , 1 ≤ index p ≤ P, 1 ≤ index q ≤ Q; signal strength sequence S p,q Including B signal strengths

[0306] The mean mass spectrum is also composed of Q first pixels; each first pixel of the mean mass spectrum corresponds one-to-one with the first pixel of each first mass spectrum corresponding to the current subset; the signal intensity sequence of each first pixel of the mean mass spectrum is denoted as S. q Each signal intensity sequence S q For the corresponding P signal intensity sequences S p,q The mean sequence; the signal strength sequence S q Including B signal strengths

[0307] It should be noted that, in this embodiment of the invention, the mean mass spectrum corresponds one-to-one with the first slice subset, and the total number of mean mass spectra is equal to the total number of sampling points N. C ;

[0308] Step 423: Assign each type of observed substance in the observed substance list as the corresponding current substance; and based on N... C Each mean mass spectrum sets the corresponding sampling data sequence for the current substance;

[0309] Specifically, this includes: step 4231, setting the mass-to-charge ratio m / z of the current substance. a As the current mass-to-charge ratio m / z * And based on the current mass-to-charge ratio m / z * Set the current mass-to-charge ratio range;

[0310] The current mass-to-charge ratio range is [m / z] * -△m / z2,m / z * +△m / z2]; △m / z2 is the preset second mass-to-charge ratio interval, △m / z2 > 0, for example, △m / z2 is 0.01Da, then the current mass-to-charge ratio interval is [m / z2]. * -0.01,m / z * +0.01];

[0311] Step 4232: Use each mean mass spectrum as the corresponding current mass spectrum;

[0312] Step 4233: The signal intensity sequences S of the current mass spectrum are... q Extract the sequence segment corresponding to the current mass-to-charge ratio interval as the current segment; and select the segment with the highest signal intensity in the current segment. As the corresponding peak intensity For the current segment excluding peak intensity The mean is calculated by taking the mean and standard deviation of all signal strengths except those mentioned above. and standard deviation And based on peak intensity mean and standard deviation Calculate the signal-to-noise ratio S / N a , And the signal-to-noise ratio (S / N) a Does it exceed the preset signal-to-noise ratio threshold S / N? hold If so, then the peak intensity will be identified. As the corresponding signal strength Otherwise, set the corresponding signal strength. The value is 0; and it is determined by the pixel coordinates and signal intensity corresponding to each first pixel point of the current mass spectrum. A corresponding second pixel is formed; and the Q obtained second pixels are combined to form a signal intensity map corresponding to the current substance;

[0313] Here, the signal-to-noise ratio threshold S / N hold For a pre-set threshold parameter, such as S / N hold =3; the signal strength map consists of Q second pixels; each second pixel includes pixel coordinates and signal strength.

[0314] Step 4234: For all non-zero signal strengths on the signal strength graph... The mean μ is obtained by calculating the mean and standard deviation. a and standard deviation σ a ; and based on the mean μ a Standard deviation σ a and signal-to-noise ratio threshold S / N hold Calculate the corresponding signal strength threshold The signal strength graph is greater than zero but less than the signal strength threshold. signal strength Reset to 0;

[0315] Step 4235: Take each second pixel of the signal strength map as the corresponding current pixel; determine the neighborhood space of the current pixel with the current pixel as the center and a preset neighborhood radius; and check all signal strengths in the current neighborhood space. Perform mean calculation and use the result as the abundance corresponding to the current pixel. And for all the obtained abundances Normalization calculations are performed to obtain the corresponding normalized abundance. And based on the pixel coordinates and normalized abundance of each second pixel point in the signal intensity map. A corresponding third pixel is formed; and the Q obtained third pixels form a normalized abundance map corresponding to the current substance.

[0316] The neighborhood radius is a pre-set radius parameter; the normalized abundance map consists of Q third pixels; each third pixel includes pixel coordinates and normalized abundance.

[0317] Step 4236: Normalize the non-zero normalized abundance values ​​on the normalized abundance map. Extract them to form the corresponding first set; and use the total abundance of the first set as the corresponding total number of samples N. i ; and the normalized abundance of each element in the first set As a corresponding matter abundance And the normalized abundance of each element in the first set The corresponding third pixel point's spatial coordinate x in d-dimensional space is used as the corresponding material coordinate. And by the coordinates of each material and their corresponding material abundance To form a corresponding sample data The corresponding time of the current mass spectrum in the continuous time period [0, T] is taken as a corresponding sampling time t. i ; and the sampling time t obtained this time i and N i Sample data Form a corresponding sampling point sample set C i ;

[0318] Step 4237, by N C The sample set C corresponding to the sampling points of the mean mass spectrum i The samples are sequentially arranged to form a sequence of sampling data corresponding to the current substance.

[0319] Step 43, and based on the sampling data sequence of various observed substances and the model training process, perform targeted training on the model parameters of the two models to obtain the corresponding first version of the parameter set;

[0320] Specifically, this includes: step 431, taking each type of observed substance as the corresponding current substance; and taking the sampling data sequence corresponding to the current substance as the current sequence;

[0321] Step 432, for the total number of samples N in the current sequence C Confirm; and extract the corresponding N from the current sequence. C Sample set of sampling points And initialize a corresponding continuous convergence counter for the current sequence to 0; and based on the initialized model parameters For the model parameters θ of the first and second models 1 θ 2 Perform initialization;

[0322] Step 433: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T Discrete time point t n ; and set the start time t of the current sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 Initialize N0 virtual particles p at each spatial point in the d-dimensional space. k ; and for each virtual particle p k Initialize a set of corresponding particle motion trajectories and mass change trajectory And each virtual particle p k coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and initialize each virtual particle p according to the mass weight initialization rule. k Quality weight Perform initialization;

[0323] Step 434: Substitute the reasoning expression of the first model into the second equation, solve the second equation using the Euler method, and complete the determination of the trajectory of each particle by solving the equation. The effective coordinates are filled; the reasoning expression of the second model is substituted into the third equation, and the third equation is solved by Euler's method. The trajectory of each mass change is completed by solving the equation. Effective weight filling;

[0324] Step 435, obtain the particle trajectories of group N0. and mass change trajectory and N C Sample set of sampling points Substitute the loss function L ALL Obtain the corresponding current loss value;

[0325] Step 436: Identify whether the current loss value meets the convergence loss value range; if not, clear the continuous convergence counter and proceed to step 437; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the continuous convergence count threshold; if it exceeds the threshold, proceed to step 438; if it does not exceed the threshold, proceed to step 437.

[0326] Step 437, based on the model optimizer, move towards making the first loss function, second loss function, third loss function and loss function L... ALL The directions that both reach their minimum values ​​are related to the model parameters θ of the first and second models. 1 θ 2 Perform one round of modulation; and return to step 434 when the current round of modulation is completed;

[0327] Step 438: Stop the current training and use the current model parameters θ of the first and second models. 1 θ 2 This forms a first-version parameter set corresponding to the current substance.

[0328] Step 5: Simulate the spatiotemporal change state of the current material based on the first set of parameters for various observed materials; identify the synthetic center region and non-center comparison region for the second slice subset of various observed materials based on the simulation data; and identify the corresponding key gene set by comparing the gene sequencing results of the synthetic center and the comparison region of various observed materials.

[0329] Specifically, this includes: Step 51, simulating the spatiotemporal change state of the current matter based on the first version of the parameter set for various observed substances;

[0330] Specifically, this includes: Step 511, dividing the continuous time period [0,T] into N equal parts according to the time interval Δt. T +1 discrete time points t n 0 ≤ index n ≤ N T ; and based on N T +1 discrete time points t n Construct a spacetime system H using d-dimensional space;

[0331] Here, it is known that the d-dimensional space includes N0 spatial points, and the coordinates of each spatial point are denoted as x. o , 1≤indexo≤N0; the spatiotemporal system H is composed of (N T +1)×N0 spacetime points Composition; each spatiotemporal point The spatiotemporal characteristics include the corresponding discrete time point t n and coordinates x o ;

[0332] Step 512: Take each type of observed substance in the observed substance list as the corresponding current substance;

[0333] Step 513: Take the first version of the parameter set corresponding to the current substance as the current parameter set; and adjust the model parameters θ of the first model and the second model based on the current parameter set. 1 θ 2 Perform the settings; and use the first and second models obtained from these settings as the corresponding current first and second models;

[0334] Step 514, from each spacetime point Discrete time point t n and coordinates x oAs a set of corresponding first times and first coordinates, a corresponding first spatiotemporal vector is input into the current first model for processing, and the first predicted velocity field obtained from this processing is used as a corresponding velocity field. And from each spacetime point Discrete time point t n and coordinates x o As a set of corresponding second time and second coordinates, a corresponding second spatiotemporal vector is input into the current second model for processing, and the first prediction-to-net generation rate obtained from this processing is used as a corresponding net generation rate.

[0335] Step 515, (N) T +1)×N0 spacetime points The corresponding discrete time point t n coordinates x o Velocity field and net generation rate Substituting the first equation yields the corresponding current partial differential equation; and solving the current partial differential equation yields each spatiotemporal point. Corresponding material density

[0336] Here, the current partial differential equation after substituting the parameters is:

[0337]

[0338] The current partial differential equations can also be solved using the Euler method, the improved Euler method, the fourth-order Runge-Kutta method, or by using publicly available partial differential equation solvers; further details will not be provided here.

[0339] Step 516, from each spacetime point The corresponding discrete time point t n coordinates x o Net generation rate and density of matter This forms a corresponding spatiotemporal feature vector; and is derived from the (N) obtained in this study. T +1)×N0 spatiotemporal point feature vectors form a first simulation data corresponding to the current substance;

[0340] Step 52, and based on the simulation data, identify the synthetic central region and non-central comparison region of the second slice subset of various observed materials;

[0341] Specifically, this includes: step 521, taking each type of observed substance in the observed substance list as the corresponding current substance;

[0342] Step 522: Take the first simulation data corresponding to the current substance as the current simulation data; and extract the corresponding (N) from the current simulation data. T +1)×N0 spatiotemporal feature vectors; and extract the corresponding discrete time point t from each spatiotemporal feature vector. n coordinates x o Net generation rate and density of matter

[0343] Step 523, based on the obtained (N) T +1)×N0 group net generation rate and density of matter The time point of strongest activity of the current substance Perform identification;

[0344] Here, the most active time point and the corresponding index n * The identification method is as follows:

[0345]

[0346] Step 524, extract the discrete time points t from the current simulation data. n With the time point of strongest activity The feature vectors of the matched spatiotemporal points are extracted to form a corresponding second set; and the metabolic flux corresponding to the feature vectors of each spatiotemporal point in the second set is calculated. Perform calculations; and calculate the resulting N0 metabolic fluxes. The mean μ is obtained by calculating the mean and standard deviation. * and standard deviation σ * ; and based on the mean μ * Standard deviation σ * Calculate the corresponding metabolic flux threshold γ hold ;

[0347] Here, metabolic flux The calculation method is as follows:

[0348]

[0349] Metabolic flux threshold γ hold The calculation method is as follows:

[0350] γ hold =μ * +β·σ * ,

[0351] Wherein, the standard deviation multiple β is a preset positive integer, for example, β = 3;

[0352] Step 525, the metabolic flux threshold γ will be greater than or equal to the threshold value. hold metabolic flux Let γ be the corresponding high-activity flux. high ; and each highly active flux γ high Corresponding coordinates x o Let x be the coordinates of the corresponding high-activity point. high And for all highly active points, the coordinates x... high All corresponding spatial points are clustered into multiple corresponding first point clusters; the total number of spatial points in each first point cluster is counted to obtain the corresponding first total number; and the first point cluster whose first total number is greater than or equal to the preset threshold of the total number of points is recorded as the corresponding second point cluster.

[0353] Here, the total number of points threshold is a pre-set integer value;

[0354] Step 526, for all highly active fluxes γ in each second cluster. high The corresponding flux sums are calculated by summing the flux sums; the second point cluster corresponding to the largest flux sum is taken as the corresponding center point cluster; and region edge identification is performed in d-dimensional space based on the center point clusters, with the spatial region of the currently identified region edge being taken as the time point of the strongest activity of the current substance. The active center region on;

[0355] Step 527: Collect the sampling time of all second slices corresponding to the current substance and the time point of strongest activity. The second closest slice subset is taken as the most active subset; and a high-resolution image of each sample slice of the most active subset is obtained by using a preset optical microscope, which is recorded as the first slice image; and all first slice images are image aligned, and all first slice images after image alignment are spatially aligned with d-dimensional space.

[0356] Here, the optical microscope in this embodiment of the invention is a high-resolution optical microscope equipped with optical slicing technology;

[0357] In all embodiments of the present invention, the image size of the first slice image is consistent; each first slice image is composed of multiple fourth pixels.

[0358] It should be noted that after image alignment, the pixel coordinates of the two fourth pixels corresponding to the same slice point on any two first slice images are the same; after spatial alignment, the two fourth pixels corresponding to the same slice point on any two first slice images correspond to the same spatial point in d-dimensional space.

[0359] Step 528: The image region corresponding to the active center region on each first slice image is recorded as the corresponding synthesis center region; and based on the preset comparison region selection rule, an image region outside the synthesis center region and with the same shape as the synthesis center region is selected on each first slice image as the corresponding non-center comparison region; and each first slice image marked with the synthesis center region and non-center comparison region is recorded as the corresponding second slice image; and all the second slice images corresponding to the current substance are combined to form a corresponding region identification data.

[0360] Here, the alignment region selection rule in this embodiment of the invention requires by default that the cell types contained in the non-center alignment region are consistent with those in the synthesis center region, and that the non-center alignment region does not contain any highly active point coordinates x in the corresponding spatial region in d-dimensional space. high Furthermore, it is required that the shape of the non-central alignment region is consistent with that of the synthetic central region;

[0361] Step 53, and identify the corresponding key gene sets by comparing the gene sequencing results of the synthesis centers and comparison regions of various observed substances;

[0362] Specifically, this includes: Step 531, taking each type of observed substance in the observed substance list as the corresponding current substance;

[0363] Step 532: Use the region identification data of the current substance as the current identification data; extract multiple second slice images marked with the synthesis center region and non-center comparison regions from the current identification data; and use the strongest active subset corresponding to the current substance as the corresponding current subset;

[0364] Here, the second slice image in this embodiment of the invention corresponds one-to-one with the sampled slices of the current subset;

[0365] Step 533: Take each second slice image as the corresponding current slice image; use a preset laser cutting microscope to align the edges of the current slice image and the corresponding sampling slice with the object; and after alignment, cut the current sampling slice into corresponding regional samples according to the synthetic center region and non-center comparison region marked on the current slice image to obtain a set of corresponding synthetic center region samples and non-center comparison region samples to form a corresponding first sample group.

[0366] Here, the scanning accuracy of the laser cutting microscope in this embodiment of the invention is by default no less than the imaging accuracy of the optical microscope;

[0367] Step 534: Based on high-throughput sequencing technology, transcriptome sequencing is performed on the synthetic central region samples and non-central alignment region samples of each first sample group to obtain corresponding central region sequencing data and alignment region sequencing data; differential gene expression analysis of the central region sequencing data and alignment region sequencing data is used to obtain the set of differentially expressed genes with an upregulated gene reading in the central region sequencing data; GO enrichment analysis of the differential gene set is used to screen the transcriptional regulatory factors of the current substance to obtain the corresponding set of first transcriptional regulatory factors; and KEGG pathway enrichment of the differential gene set is used to screen the metabolic enzyme genes of the synthetic transformation pathway of the current substance to obtain the corresponding set of first metabolic enzyme genes.

[0368] Step 535: Take the intersection of the first transcription regulatory factor sets of all first sample groups to obtain the corresponding second transcription regulatory factor sets; and take the intersection of the first metabolic enzyme gene sets of all first sample groups to obtain the corresponding second metabolic enzyme gene sets; and form the corresponding key gene set by the second transcription regulatory factor sets and the second metabolic enzyme gene sets.

[0369] Step 6: Construct corresponding substance analysis reports based on the synthesis centers and key gene analysis information of various observed substances; and combine all substance analysis reports to form a metabolite analysis report corresponding to the target substance and save it.

[0370] Specifically, this includes: Step 61, constructing corresponding material analysis reports based on the synthesis centers and key gene analysis information of various observed materials;

[0371] Specifically, this includes: taking each type of observed substance in the observation substance list as the corresponding current substance; and forming a corresponding substance analysis report consisting of the strongest activity time point corresponding to the current substance, as well as all second slice images and key gene sets corresponding to the strongest activity time point;

[0372] Step 62, and then compile the analysis report of the metabolites corresponding to the target substance from all the substance analysis reports and save it.

[0373] Here, when generating metabolite analysis reports, all substance analysis reports can be sorted according to the chronological order of the time points of the strongest activity of each type of observed substance.

[0374] Figure 2 This is a modular structure diagram of an analysis device for plant metabolite synthesis centers and key genes provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the aforementioned method embodiments, or it can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiments. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2As shown, the device includes: an analysis tool preparation module 201, an experimental material preparation module 202, an observation and sampling module 203, an adaptive training module 204, a simulation analysis module 205, and a report generation module 206.

[0375] The analysis tool preparation module 201 is used to construct partial differential equations to describe how the state of matter changes based on the spatiotemporal changes of the state of matter transportation and the state of matter synthesis or consumption, and to denote the current partial differential equation as the first equation; to design corresponding neural network models for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation, and to denote the corresponding first model and second model; to set two corresponding ordinary differential equations for the first equation, and to denote the corresponding second equation and third equation; to customize the data format for the sampled data sequence loaded with experimental sampling data; and to design the corresponding loss function L. ALL Based on the sampled data sequence, the second equation, the third process, and the loss function L... ALL Design a model training process for training the first and second models.

[0376] The experimental material preparation module 202 is used to extract the synthesis site of the target substance from the plant material as the experimental material; to select an experimental reagent to spray the experimental material; and to store the treated experimental material in a preset humidification device.

[0377] The observation and sampling module 203 is used to take the material preservation start time as the start observation time and customize the sampling time axis based on the preset observation duration and sampling time interval; and to sample the experimental material once every sampling time interval from the start observation time based on the preset frozen section sampling rules to obtain the corresponding sampling section set; and to divide the current sampling section set into the first section subset and the second section subset and store them in the preset freezing device.

[0378] The adaptive training module 204 is used to construct an observation material list based on all upstream and downstream metabolites along the target material synthesis pathway after sampling; extract sampling data of various observation materials in the observation material list based on all first slice subsets to obtain corresponding sampling data sequences; and perform targeted training on the model parameters of the two models based on the sampling data sequences of various observation materials and the model training process to obtain the corresponding first version of parameter set.

[0379] The simulation analysis module 205 simulates the spatiotemporal change state of the current material based on the first set of parameters for various observed materials; and identifies the synthetic center region and non-center comparison region for the second slice subset of various observed materials based on the simulation data; and identifies the corresponding key gene set by comparing the gene sequencing results of the synthetic center and the comparison region of various observed materials.

[0380] The report generation module 206 constructs corresponding material analysis reports based on the synthesis centers and key gene analysis information of various observed substances; and the metabolite analysis report corresponding to the target substance is composed of all the material analysis reports and saved.

[0381] The present invention provides an analytical device for plant metabolite synthesis centers and key genes, which can perform the method steps in the above method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0382] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, the analysis tool preparation module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0383] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0384] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the foregoing method embodiments are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0385] Figure 3 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be a terminal device or server implementing the methods of the aforementioned embodiments, or it can be a terminal device or server connected to the aforementioned terminal device or server implementing the methods of the aforementioned embodiments. Figure 3 As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0386] exist Figure 3The system bus 305 mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0387] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0388] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

[0389] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for analyzing plant metabolite synthesis centers and key genes. As described above, this invention pre-constructs a partial differential equation (i.e., the first equation) to describe how the state of matter changes based on the spatiotemporal variations of the state of matter transport and the state of matter synthesis or consumption. Two artificial intelligence models (i.e., the first model and the second model) are designed for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation. Driven by the velocity field spatiotemporal function and the net generation rate spatiotemporal function of the first equation, two ordinary differential equations (the second equation and the third equation) are set to simulate the coordinates and mass weights of virtual particles from a Lagrange perspective. The solution process of the two ordinary differential equations is combined to design the corresponding loss function L for the joint training of the first and second models. ALL Based on the sampled data sequence, two ordinary differential equations, and the loss function L... ALLA standard processing flow for joint training (i.e., model training flow) is designed. This embodiment of the invention obtains multiple sampling slice sets at various observation times through experimental observation, and divides each sampling slice set into two subsets (first slice subset and second slice subset). The first slice subset at all observation times is used for the analysis of the strongest activity time point and the location of the synthesis center, while the second slice subset closest to the strongest activity time point is used for key gene information analysis. In analyzing the strongest activity time point and the location of the synthesis center, this embodiment first constructs a list of observed substances based on all known upstream and downstream metabolites along the target substance's synthesis pathway. Then, based on all first slice subsets, the sampling data sequences of various observed substances are constructed. Next, based on the sampling data sequences of various observed substances and the standard model training flow, the model parameters of the two models are trained specifically to obtain a set of model parameters corresponding to the current substance (i.e., the first set of parameters). Finally, based on the model parameter sets of various observed substances, spatiotemporal change state simulations are performed, and the strongest activity time point and the location of the synthesis center for each observed substance are confirmed by analyzing the spatiotemporal simulation data. In analyzing key gene information of various observed substances, this invention selects the second subset of slices closest to the sampling time and the time point of strongest activity as the analysis subset for the current substance. Central / alignment region samples are extracted from the sampled slices in the current analysis subset using optical and laser cutting microscopes. Transcriptome sequencing is then performed on the central / alignment region samples, and the corresponding key gene information is identified through alignment sequencing results. This invention achieves the technical objective of locating the time point of strongest activity and the center of synthesis for each type of metabolite in a spatiotemporal system. The adaptive model training process corresponding to each type of observed substance improves model generalization and localization accuracy. Furthermore, this invention improves the precision and accuracy of central / alignment region sample extraction, thereby enhancing the accuracy of key gene information identification.

[0390] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0391] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for analyzing the biosynthetic centers and key genes of plant metabolites, characterized in that, The method includes: This paper constructs partial differential equations to describe how the state of matter changes based on the spatiotemporal variations of the state of matter transportation and the state of matter synthesis or consumption. This partial differential equation is denoted as the first equation. Corresponding neural network models are designed for the spatiotemporal functions of the velocity field and net generation rate in the first equation, denoted as the first model and the second model. Two corresponding ordinary differential equations are set for the first equation, denoted as the second equation and the third equation. A customized data format is provided for the sampled data sequence loaded with experimental data, and a corresponding loss function L is designed. ALL Based on the sampled data sequence, the second equation, the third-party process, and the loss function L... ALL Design a model training process for training the first model and the second model; The synthesis site of the target substance on plant material was extracted as experimental material; an experimental reagent was selected to spray the experimental material; and the treated experimental material was stored in a pre-set humidification device. The material preservation start time is used as the start observation time, and a sampling time axis is customized based on the preset observation duration and sampling time interval. Starting from the start observation time, the experimental material is sampled once every sampling time interval based on the preset frozen section sampling rules to obtain the corresponding sampled section set. The current sampled section set is divided into a first section subset and a second section subset and stored in a preset freezing device. After sampling, an observation substance list is constructed based on all upstream and downstream metabolites along the target substance synthesis pathway; and sampling data is extracted from each type of observation substance in the observation substance list based on all the first slice subsets to obtain the corresponding sampling data sequence; and the model parameters of the two models are trained in a targeted manner based on the sampling data sequence of each type of observation substance and the model training process to obtain the corresponding first version of the parameter set. The spatiotemporal change state of the current material is simulated based on the first set of parameters of various observed materials; and the synthetic center region and non-center comparison region are identified in the second slice subset of various observed materials based on the simulation data; and the corresponding key gene set is identified by comparing the gene sequencing results of the synthetic center and the comparison region of various observed materials. Based on the analysis information of the synthesis centers and key genes of the various observed substances, corresponding substance analysis reports are constructed; and the metabolite analysis reports corresponding to the target substance are composed of all the substance analysis reports and saved.

2. The method for analyzing plant metabolite synthesis centers and key genes according to claim 1, characterized in that, The first equation is: Where t is time and x is the coordinate of a point in space; The time t is a point in time within a preset continuous time period [0, T], and T is a preset time length; The spatial point coordinate x is the coordinate of a point in a preset d-dimensional space, where d is 2 or 3; the d-dimensional space is a spatial point network composed of multiple evenly distributed spatial points, each spatial point corresponding to a d-dimensional point coordinate; the total number of spatial points in the d-dimensional space is denoted as N0. ρ(t,x) is a density-spacetime function with the dimension of mass / space, used to describe the density of the corresponding spatial point at time t at the coordinate x of the spatial point. v(t,x) is a spacetime function of the velocity field with the dimension of space / time. It is used to describe the d-dimensional velocity field of the corresponding spatial point at time t, where the d-dimensional velocity field is composed of d-dimensional material outflow velocity components. If each component is greater than 0, it indicates a net outflow of material in the current dimension; if it is less than 0, it indicates a net inflow of material in the current dimension. g(t,x) is a spacetime function of the net generation rate with the dimension 1 / time. It is used to describe the net generation rate of the corresponding spatial point at time t, where the net generation rate represents the proportion of the change in the amount of matter caused by the synthesis or consumption of local matter at the current spatial point per unit time. This is the local time derivative term of the partial differential equation, used to describe the rate of change of density of the corresponding spatial point at time t at the coordinate x of the spatial point. Here, v(t,x)·ρ(t,x) is the divergence term of the partial differential equation, used to describe the mass flux at time t at the spatial point corresponding to x. Used to describe the spatial divergence of mass flux; a spatial divergence greater than 0 indicates a net outflow of mass, and a spatial divergence less than 0 indicates a net inflow of mass. g(t,x)·ρ(t,x) are the source and sink terms of the partial differential equation, where g(t,x) is the source term and ρ(t,x) is the sink term. The source and sink terms are used to describe the density change of the spatial point corresponding to x at time t caused by the synthesis or consumption of local matter at the spatial point, without considering the factors of matter flow. The first model and the second model are each implemented based on a type of neural network model used for processing regression calculation tasks; the neural network model used for processing regression calculation tasks includes MLP model and CNN model; The first model is used to predict the velocity field based on the first spatiotemporal vector input to the model and output the corresponding first predicted velocity field; the first spatiotemporal vector includes a first time and a first coordinate; the first time and the first coordinate correspond to the time t and the spatial point coordinate x of the velocity field spatiotemporal function v(t,x), respectively; the first predicted velocity field includes d first predicted velocity components; the model inference expression of the first model is: v * =F 1 (t,x;θ 1 );v * For the first predicted velocity field, F 1 () represents the reasoning process of the first model, θ 1 These are the model parameters of the first model; The second model is used to predict the net generation rate based on the second spatiotemporal vector input to the model and output the corresponding first predicted net generation rate; the second spatiotemporal vector includes a second time and a second coordinate; the second time and the second coordinate correspond to the time t and the spatial point coordinate x of the net generation rate spatiotemporal function g(t,x), respectively; the model inference expression of the second model is: g * =F 2 (t,x;θ 2 ); where g * F is the first prediction ratio to net generation rate. 2 () represents the reasoning process of the second model, θ 2 These are the model parameters for the second model; The observation duration is an integer multiple of the sampling time interval, and the duration T = observation duration - sampling time interval; the starting zero time of the continuous time period [0, T] corresponds to the first sampling point on the sampling time axis, and the time interval between the first sampling point and the starting observation time is the sampling time interval; the ending time of the continuous time period [0, T] corresponds to the last sampling point on the sampling time axis, and the last sampling point is the ending time of the sampling time axis; the total number of sampling points N on the sampling time axis C = Observation duration / Sampling time interval; The data format of the sampling data sequence is as follows: the sampling data sequence consists of a sample set C of multiple sampling points. i Composition, 0 ≤ index i ≤ N C ; Sample set C for each sampling point i Including a sampling time t i and multiple sample data The sampling time t i For a given moment in the continuous time interval [0, T], 1 ≤ index j ≤ N i N i The sample set C of the sampling points i The total number of samples; the sample data Including material coordinates Material abundance 3. The method for analyzing plant metabolite synthesis centers and key genes according to claim 2, characterized in that, The step of setting two corresponding ordinary differential equations for the first equation, denoted as the corresponding second equation and third equation, specifically includes: Step 31: Based on the Lagrange perspective, construct N0 virtual particles p in the d-dimensional space. k , 1≤index k≤N0; and for each of the virtual particles p k Set the corresponding coordinates and quality weight And each of the virtual particles p k Coordinates of the starting time of the continuous time interval [0,T] Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and initialize each virtual particle p according to the preset mass weight initialization rule. k The mass weight at the start time of the continuous time interval [0,T] Perform initialization; Wherein, the virtual particle p k The starting time of the continuous time interval [0, T] corresponds one-to-one with a spatial point in the d-dimensional space; the coordinates The mass weights are the real-time coordinates of the k-th virtual particle in the d-dimensional space at time t. Let be the total amount of matter carried by the k-th virtual particle at time t; Step 32: Based on the first equation, design two ordinary differential equations for each virtual particle p. k The coordinates at any time t within the continuous time period [0,T] and the quality weight Evolution; The two ordinary differential equations include the second equation and the third equation; The second equation is a flow mapping equation for the time-varying coordinates of a virtual particle driven by the spacetime function v(t,x) of the velocity field, specifically: The solution to the second equation is For any virtual particle p, the time evolution mapping is used. k In terms of time evolution mapping Corresponding to a dataset This dataset contains the particle trajectory of the current virtual particle. Substituting the first model inference expression corresponding to the velocity field spacetime function v(t,x) into the second equation, we obtain: The third process is a flow mapping equation for the evolution of the mass weight of virtual particles over time, driven by the spatiotemporal function g(t,x) of the net generation rate, specifically: The solution of the third-party process is For any virtual particle p, the time evolution mapping of the mass weights is defined. k In terms of time evolution mapping Corresponding to a dataset This dataset represents the trajectory of the current virtual particle's mass change. Substituting the second model inference expression corresponding to the net generation rate spatiotemporal function g(t,x) into the third process, we obtain:

4. The method for analyzing plant metabolite synthesis centers and key genes according to claim 3, characterized in that, The process of solving the second equation using the Euler method is as follows: The continuous time period [0, T] is divided into multiple discrete time points t according to a preset time interval Δt. n 0 ≤ index n ≤ N T The total number of discrete points N T = T / Δt; the time interval Δt is less than the sampling time interval; and the second equation Approximately: When n=0, The virtual particle p k coordinates Given a quantity, then: When n=1, The virtual particle p is obtained by iteratively obtaining the virtual particle p point by point. k N in the continuous time period [0,T] T +1 coordinates Composed of corresponding particle trajectories The particle trajectory Given the given information, the process of solving the third equation using the Euler method is as follows: The third-party program Approximately: When n=0, The virtual particle p k Quality weight Given a quantity, then: When n=1, The virtual particle p is obtained by iteratively obtaining the virtual particle p point by point. k N in the continuous time period [0,T] T +1 quality weight The corresponding mass change trajectory 5. The method for analyzing plant metabolite synthesis centers and key genes according to claim 4, characterized in that, The loss function L ALL For: L ALL =λ1L1+λ2L2+λ3L3; Where λ1, λ2, and λ3 are the preset first, second, and third loss weights, and L1, L2, and L3 are the corresponding first, second, and third loss functions; The first loss function L1 is a regularization loss, specifically: Where, λ v , λ g These are two learnable weight parameters; The second loss function L2 is a quality loss, specifically: in, For all virtual particles at the sampling time t i The predicted total amount of matter above, The sampling time t is calculated based on the sampled data sequence. i The total amount of matter on the surface For the first sampling time t i＝1 Corresponding observation; f m→w () is the preset abundance-quality weighting conversion function; Quality weight The calculation method is as follows: For the known mass change trajectory Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n The corresponding quality weight As the current quality weight If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let time t be denoted as t - t + , t - <t i <t + And based on the time t before and after - t + The two corresponding quality weights The current quality weight is obtained by performing linear interpolation. The third loss function L3 is an optimal transmission loss, specifically: Among them, F dis () represents the preset Wasserstein distance function; The total mass of all virtual particles at the sampling time t i Discrete distribution predictor on, The sampling time t is calculated based on the sampled data sequence. i Discrete distribution observations on; δ is the Dirac measure, Dirac measure coordinates The unit point mass of the corresponding spatial point, Dirac measure coordinates The unit point mass of the corresponding spatial point; coordinate The calculation method is as follows: For the known particle trajectory Perform a query; if there exists a time t n With the current time t i If they are equal, then time t will be... n The corresponding coordinates As current coordinates If there is no time t i Equal time t n Then it will be related to the current time t i The two closest moments t n Let these be time points t- and t-. + , t - <t i <t + And based on the time intervals t- and t-before, + The two corresponding coordinates Perform d-dimensional coordinate interpolation calculation and use the calculation result as the current coordinates.

6. The method for analyzing plant metabolite synthesis centers and key genes according to claim 5, characterized in that, The method based on the sampled data sequence, the second equation, the third procedure, and the loss function L ALL Design a model training process for training the first model and the second model, specifically including: Step 61, for the total number of samples N of the current sampling data sequence C Confirm; and extract the corresponding N from the sequence. C The sample set of the sampling points The continuous convergence counter is initialized to 0; and a set of preset initialization model parameters are used. The model parameters θ of the first model and the second model 1 θ 2 Perform initialization; Step 62: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T The discrete time points t n ; and the start time t of the sampled data sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 At each spatial point location in the d-dimensional space, N0 virtual particles p are initialized. k ; and for each of the virtual particles p k Initialize a set of corresponding particle motion trajectories and the mass change trajectory And each of the virtual particles p k The coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and apply the mass weight initialization rule to each virtual particle p. k Quality weight Perform initialization; The initialization rule for the quality weights is as follows: Each of the quality weights... As the current weight, and the coordinates corresponding to the current weight are used. Used as the current coordinates; and for the sample set of sampling points Perform a query; if a given material coordinate exists... If the coordinates are the same as the current coordinates, then the abundance-quality weighting transformation function f is used. m→w () for the current material coordinates The corresponding abundance m of the substance j i Perform quality weight transformation and set the current weight based on the transformation result. If there are no material coordinates that match the current coordinates Then based on the preset default quality weight w default Set the current weight. The motion trajectories of each particle after initialization In addition to the first coordinate mentioned All coordinates other than the valid coordinates are invalid coordinates; the mass change trajectories after initialization are all invalid coordinates. In addition to the first mentioned quality weight All weights other than the valid weights are invalid weights. Step 63: Substitute the first model inference expression into the second equation, solve the second equation using the Euler method, and complete the determination of the motion trajectory of each particle by solving the equation. The effective coordinates are filled; the second model inference expression is substituted into the third equation, and the third equation is solved using the Euler method. The mass change trajectories are then solved by solving the equations. Effective weight filling; Step 64, obtain the particle motion trajectories of group N0. and the mass change trajectory and N C The sample set of the sampling points Substitute the loss function L ALL Obtain the corresponding current loss value; Step 65: Identify whether the current loss value meets the preset convergence loss value range; if not, clear the continuous convergence counter and proceed to step 66; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the preset continuous convergence count threshold; if it exceeds the threshold, proceed to step 67; if it does not exceed the threshold, proceed to step 66. Step 66: Based on a preset model optimizer, optimize the first loss function, the second loss function, the third loss function, and the loss function L. ALL The direction in which both reach their minimum values ​​is related to the model parameters θ of the first and second models. 1 θ 2 Perform one round of modulation; and return to step 63 when the current round of modulation is completed; The model optimizer includes the Adam optimizer and the SGD optimizer; Step 67: Stop the current training and use the current model parameters θ of the first model and the second model. 1 θ 2 This forms a corresponding first version parameter set.

7. The method for analyzing plant metabolite synthesis centers and key genes according to claim 1, characterized in that, The target substance is a class of terpenoid compounds; The target substance on the plant material is synthesized from plant branches, plant roots, or plant stems. The experimental reagents are a class of reagents that can stimulate the metabolic activity of target substances, including jasmonic acid reagent, methyl jasmonic acid reagent, crotonin reagent, and modified jasmonic acid derivatives. The humidification device is a type of enclosed device equipped with an inlet / outlet air pump, a temperature and humidity sensor, a humidification device, and a temperature control device. The humidification device is used to regulate the ventilation state inside the device by adjusting the inlet / outlet air pump, monitor the temperature and humidity state inside the device by the temperature and humidity sensor, regulate the humidity state inside the device by the humidification device, and regulate the temperature state inside the device by the temperature control device. The freezing device is a type of ultra-low temperature freezing equipment, with a freezing temperature that can reach -80℃; The list of observed substances includes multiple categories of observed substances; each category of observed substances is an upstream or downstream metabolite in the synthetic pathway of the target substance.

8. The method for analyzing plant metabolite synthesis centers and key genes according to claim 1, characterized in that, The step of sampling the experimental material once every sampling time interval from the start observation time to obtain a corresponding sample slice set based on a preset frozen section sampling rule specifically includes: During the first sampling, the experimental material is removed from the humidification device; the tail end or head end of the experimental material is selected as the initial starting position; pruning shears are used to cut off and discard branches, roots, or stems approximately 0.5 cm after the initial starting position, and the current cut position of the experimental material is taken as the starting position of the current sampling; pruning shears are used again to cut off branches, roots, or stems approximately 0.5 cm after the initial sampling position, and the cut branches, roots, or stems are taken as the sampling material for this sampling; the previous cut position is marked on the current cut position of the experimental material; the marked experimental material is returned to the humidification device; and cryosectioning technology is used to continuously slice the sampling material according to a preset single slice thickness to obtain multiple corresponding sampling slices to form the corresponding sampling slice set; the thickness of each single slice is not less than 10 μm. In each subsequent sampling, the experimental material is removed from the humidification device; the previously marked cutting position is taken as the starting position for the current sampling; pruning shears are used to cut and discard branches, roots, or stems approximately 0.5 cm after the current starting position, and the current cutting position of the experimental material is taken as the starting position for the current sampling; pruning shears are used again to cut branches, roots, or stems approximately 0.5 cm after the current sampling starting position, and the cut branches, roots, or stems are taken as the current sampling material; the previous cutting position is marked on the current cutting position of the experimental material; the marked experimental material is returned to the humidification device; and cryosectioning technology is used to continuously slice the current sampling material according to the thickness of a single slice to obtain multiple corresponding sampling slices to form the corresponding sampling slice set.

9. The method for analyzing plant metabolite synthesis centers and key genes according to claim 2, characterized in that, The step of extracting the corresponding sampling data sequence from various types of observed substances in the observed substance list based on all the first slice subsets specifically includes: Step 91: Set a preset mass-to-charge ratio sequence for the mass spectrometry imaging device based on the observed substance list; and use the mass spectrometry imaging device to perform spatial mass spectrometry imaging on each sampling slice of each first slice subset based on the mass-to-charge ratio sequence to obtain the corresponding first mass spectrum; and perform image alignment on all the first mass spectra, and spatially align all the image-aligned first mass spectra with the d-dimensional space of the first equation. The mass spectrometry imaging device is a analytical ionization mass spectrometry imaging device. The total number of observed substances in the observed substance list is denoted as A; each category of observed substances corresponds to a mass-to-charge ratio m / z. a 1 ≤ index a ≤ A; the total number of mass-to-charge ratios in the mass-to-charge ratio sequence is denoted as B, consisting of multiple mass-to-charge ratios m / z. b The order is determined by sequential sorting, where 1 ≤ index b ≤ B, and the mass-to-charge ratio m / z is between every two adjacent elements. b The difference between them is a preset first mass-to-charge ratio interval Δm / z1; the mass-to-charge ratio sequence covers all the mass-to-charge ratios m / z. a ; All the first mass spectra have the same image size; each first mass spectra consists of multiple first pixels; each first pixel includes pixel coordinates, the mass-to-charge ratio sequence, and a signal intensity sequence S; the pixel coordinates are two-dimensional or three-dimensional coordinates; the signal intensity sequence S consists of B signal intensities s. b The signal strength s is ordered sequentially. b With the mass-to-charge ratio m / z b One-to-one correspondence; After image alignment, the pixel coordinates of the two first pixels corresponding to the same slice point on any two first mass spectra are the same, wherein the same slice point refers to the point on each slice that has the same relative position with respect to the slice's central axis; after spatial alignment, the two first pixels corresponding to the same slice point on any two first mass spectra correspond to the same spatial point in the d-dimensional space. Step 92: Take each of the first slice subsets as the corresponding current subset; and calculate the mean signal intensity of all the first mass spectra in the current subset to obtain the corresponding mean mass spectra; The total number of mass spectra of the current subset is denoted as P; The total number of pixels in each of the first mass spectra is equal, denoted as the total number Q; Each of the first mass spectra consists of Q first pixels; the signal intensity sequence of each first pixel in each first mass spectra is denoted as S. p,q 1 ≤ index p ≤ P, 1 ≤ index q ≤ Q; the signal strength sequence S p,q Including B signal strengths The mean mass spectrum is also composed of Q first pixels; the first pixels of the mean mass spectrum correspond one-to-one with the first pixels of each first mass spectrum corresponding to the current subset; the signal intensity sequence of each first pixel of the mean mass spectrum is denoted as S. q Each of the signal strength sequences S q For the corresponding P signal intensity sequences S p,q The mean sequence; the signal strength sequence S q Including B signal strengths The mean mass spectrum corresponds one-to-one with the first slice subset, and the total number of the mean mass spectra is equal to the total number of sampling points N. C ; Step 93: Take each type of observed substance in the observed substance list as the corresponding current substance; and based on N... C The mean mass spectrum is the sampling data sequence corresponding to the current substance.

10. The method for analyzing plant metabolite synthesis centers and key genes according to claim 9, characterized in that, The N-based C The mean mass spectrum is the sampling data sequence corresponding to the current substance, specifically including: Step 101, set the mass-to-charge ratio m / z corresponding to the current substance. a As the current mass-to-charge ratio m / z * And based on the current mass-to-charge ratio m / z * Set the current mass-to-charge ratio range; Wherein, the current mass-to-charge ratio range is [m / z * -△m / z2,m / z * +△m / z2];△m / z2 is the preset second mass-to-charge ratio interval, △m / z2>0; Step 102: Use each of the mean mass spectra as the corresponding current mass spectra; Step 103, convert each of the signal intensity sequences S of the current mass spectrum. q The sequence segment corresponding to the current mass-to-charge ratio interval is extracted as the current segment; and the maximum signal intensity in the current segment is selected. As the corresponding peak intensity For the current segment excluding the peak intensity The mean is calculated by taking the mean and standard deviation of all signal strengths except those mentioned above. and standard deviation And based on the peak intensity The mean and the standard deviation Calculate the signal-to-noise ratio S / N a , Regarding the signal-to-noise ratio S / N a Does it exceed the preset signal-to-noise ratio threshold S / N? hold If the peak intensity is detected, then the peak intensity will be... As the corresponding signal strength Otherwise, set the corresponding signal strength. =0; and determined by the pixel coordinates corresponding to each of the first pixels in the current mass spectrum and the signal intensity. A corresponding second pixel is formed; and the Q second pixels are combined to form a signal intensity map corresponding to the current substance. The signal strength map is composed of Q second pixels; each second pixel includes the pixel coordinates and the signal strength. Step 104: For all non-zero signal strengths on the signal strength graph... The mean μ is obtained by calculating the mean and standard deviation. a and standard deviation σ a ; and based on the mean μ a The standard deviation σ a and the signal-to-noise ratio threshold S / N hold Calculate the corresponding signal strength threshold The signal strength graph is greater than zero but less than the signal strength threshold. The signal strength Reset to 0; Step 105: Take each of the second pixels in the signal strength map as the corresponding current pixel; and determine the neighborhood space of the current pixel with the current pixel as the center and a preset neighborhood radius; and check all the signal strengths in the current neighborhood space. Perform mean calculation and use the result as the abundance corresponding to the current pixel. And for all the abundances obtained Normalization calculations are performed to obtain the corresponding normalized abundance. And the pixel coordinates corresponding to each of the second pixels in the signal intensity map and the normalized abundance. A corresponding third pixel is formed; and the Q third pixels are combined to form a normalized abundance map corresponding to the current substance. The normalized abundance map is composed of Q third pixels; each third pixel includes the pixel coordinates and the normalized abundance. Step 106: The non-zero normalized abundance values ​​on the normalized abundance map are... Extract them to form the corresponding first set; and use the total abundance of the first set as the corresponding total number of samples N. i ; and the normalized abundance of each of the first set As a corresponding substance abundance and the normalized abundance of each of the first set The spatial coordinate x of the corresponding spatial point of the third pixel in the d-dimensional space is used as the corresponding material coordinate. And by the coordinates of each of the aforementioned substances and the corresponding abundance of the substances To form a corresponding sample data The current mass spectrum is used as the corresponding sampling time t within the continuous time period [0, T]. i ; and from the sampling time t obtained this time i and N i The sample data Form a corresponding sample set C of the sampling points i ; Step 107, by N C The sample set C corresponding to the mean mass spectrum. i The samples are sequentially arranged to form a sequence of sampling data corresponding to the current substance.

11. The method for analyzing plant metabolite synthesis centers and key genes according to claim 6, characterized in that, The first set of parameters is obtained by performing targeted training on the model parameters of the two models based on the sampling data sequences of various observed substances and the model training process, specifically including: Step 111: Take each type of observed substance as the corresponding current substance; and take the sampling data sequence corresponding to the current substance as the current sequence; Step 112, for the total number of samples N of the current sequence C Confirm; and extract the corresponding N from the current sequence. C The sample set of the sampling points And initialize a corresponding continuous convergence counter for the current sequence to 0; and based on the initialized model parameters The model parameters θ of the first model and the second model 1 θ 2 Perform initialization; Step 113: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T The discrete time points t n ; and the start time t of the current sequence i＝1 With discrete time point t n＝0 Alignment; and based on the Lagrange perspective, at time t n＝0 At each spatial point location in the d-dimensional space, N0 virtual particles p are initialized. k ; and for each of the virtual particles p k Initialize a set of corresponding particle motion trajectories and the mass change trajectory And each of the virtual particles p k The coordinates Initialize the virtual particle p to the point coordinates of the corresponding spatial point; and apply the mass weight initialization rule to each virtual particle p. k Quality weight Perform initialization; Step 114: Substitute the first model inference expression into the second equation, solve the second equation using the Euler method, and complete the determination of the motion trajectory of each particle by solving the equation. The effective coordinates are filled; the second model inference expression is substituted into the third equation, and the third equation is solved using the Euler method. The mass change trajectories are then solved by solving the equations. Effective weight filling; Step 115, obtain the particle motion trajectories of group N0. and the mass change trajectory and N C The sample set of the sampling points Substitute the loss function L ALL Obtain the corresponding current loss value; Step 116: Identify whether the current loss value meets the convergence loss value range; if not, clear the continuous convergence counter and proceed to step 117; if it meets the range, increment the continuous convergence counter by 1, and identify whether the incremented continuous convergence counter exceeds the continuous convergence count threshold; if it exceeds, proceed to step 118; if it does not exceed, proceed to step 117. Step 117, based on the model optimizer, move towards making the first loss function, the second loss function, the third loss function, and the loss function L... ALL The direction in which both reach their minimum values ​​is related to the model parameters θ of the first and second models. 1 θ 2 Perform one round of modulation; and return to step 114 when the current round of modulation is completed; Step 118: Stop the current training and use the current model parameters θ of the first model and the second model. 1 θ 2 Form a first version parameter set corresponding to the current substance.

12. The method for analyzing plant metabolite synthesis centers and key genes according to claim 3, characterized in that, The simulation of the spatiotemporal change state of the current matter based on the first set of parameters for various types of observed matter specifically includes: Step 121: Divide the continuous time period [0,T] into N equal parts according to the time interval Δt. T +1 of the discrete time points t n 0 ≤ index n ≤ N T ; and based on N T +1 of the discrete time points t n Construct a spacetime system H with the d-dimensional space; The d-dimensional space includes N0 spatial points, and the coordinates of each spatial point are denoted as x. o , 1 ≤ index o ≤ N0; the spatiotemporal system H is composed of (N T +1)×N0 spacetime points Composition; each of the aforementioned spatiotemporal points The spatiotemporal characteristics include the corresponding discrete time point t n and the coordinate x o ; Step 122: Take each type of observed substance in the observed substance list as the corresponding current substance; Step 123: Take the first version of the parameter set corresponding to the current substance as the corresponding current parameter set; and adjust the model parameters θ of the first model and the second model based on the current parameter set. 1 θ 2 Perform the settings; and use the first model and the second model obtained from this setting as the corresponding current first model and current second model; Step 124, from each of the aforementioned spacetime points The discrete time point t n and the coordinate x o As a set of corresponding first times and first coordinates, a corresponding first spatiotemporal vector is input into the current first model for processing, and the first predicted velocity field obtained from this processing is used as a corresponding velocity field. And by each of the aforementioned spacetime points The discrete time point t n and the coordinate x o As a set of corresponding second times and second coordinates, a corresponding second spatiotemporal vector is input into the current second model for processing, and the first predicted net generation rate obtained from this processing is used as a corresponding net generation rate. Step 125, (N) T +1)×N0 of the aforementioned spacetime points The corresponding discrete time point t n The coordinate x o The velocity field and the net generation rate Substituting the first equation yields the corresponding current partial differential equation; and solving the current partial differential equation yields each of the spatiotemporal points. Corresponding material density The current partial differential equation is: Step 126, from each of the aforementioned spacetime points The corresponding discrete time point t n The coordinate x o The net production rate and the density of the substance This forms a corresponding spatiotemporal feature vector; and is derived from the (N) obtained in this study. T +1)×N0 of the spatiotemporal point feature vectors form a first simulation data corresponding to the current substance.

13. The method for analyzing plant metabolite synthesis centers and key genes according to claim 12, characterized in that, The process of identifying the synthetic central region and non-central comparison regions of the second slice subset of various observed materials based on simulated data specifically includes: Step 131: Take each type of observed substance in the observed substance list as the corresponding current substance; Step 132: Take the first simulation data corresponding to the current substance as the current simulation data; and extract the corresponding (N) from the current simulation data. T +1)×N0 spatiotemporal point feature vectors; and extract the corresponding discrete time point t from each of the spatiotemporal point feature vectors. n The coordinate x o The net production rate and the density of the substance Step 133, based on the obtained (N) T +1)×N0 group's net generation rate and the density of the substance The time point of strongest activity of the current substance Perform identification; Among them, the strongest activity time point and the corresponding index n * The identification method is as follows: Step 134, the discrete time point t in the current simulation data n With the strongest activity time point The matched spatiotemporal point feature vectors are extracted to form a corresponding second set; and the metabolic flux corresponding to each spatiotemporal point feature vector in the second set is... Perform calculations; and for the obtained N0 metabolic fluxes... The mean μ is obtained by calculating the mean and standard deviation. * and standard deviation σ * ; and based on the mean μ * The standard deviation σ * Calculate the corresponding metabolic flux threshold γ hold ; Wherein, the metabolic flux The calculation method is as follows: The metabolic flux threshold γ hold The calculation method is as follows: c hold =μ * +b·s * , The standard deviation factor β is a preset positive integer; Step 135, the metabolic flux threshold γ is greater than or equal to the threshold value. hold The metabolic flux Let γ be the corresponding high-activity flux. high ; and each of the aforementioned high-activity flux γ high The corresponding coordinate x o Let x be the coordinates of the corresponding high-activity point. high ; and for all the aforementioned highly active point coordinates x high All corresponding spatial points are clustered into multiple corresponding first point clusters; the total number of spatial points in each first point cluster is counted to obtain the corresponding first total number; and the first point clusters whose first total number is greater than or equal to a preset threshold for the total number of points are recorded as the corresponding second point clusters. Step 136, for all the highly active fluxes γ in each of the second point clusters high The corresponding flux sum is obtained by summing the flux sums; the second point cluster corresponding to the largest flux sum is taken as the corresponding center point cluster; and region edge identification is performed in the d-dimensional space based on the center point cluster, and the spatial region of the currently identified region edge is taken as the current material at the time point of strongest activity. The active center region on; Step 137: Combine the sampling times of all the second slices corresponding to the current substance with the time point of strongest activity. The closest second slice subset is taken as the most active subset; and a high-resolution image of each sample slice of the most active subset is obtained by using a preset optical microscope, which is recorded as the first slice image; and all the first slice images are image aligned, and all the image-aligned first slice images are spatially aligned with the d-dimensional space. The optical microscope is a high-resolution optical microscope equipped with optical slicing technology. All the first slice images have the same image size; each first slice image consists of multiple fourth pixels; After image alignment, the pixel coordinates of the two fourth pixels corresponding to the same slice point on any two first slice images are the same; after spatial alignment, the two fourth pixels corresponding to the same slice point on any two first slice images correspond to the same spatial point in the d-dimensional space. Step 138: The image region corresponding to the active center region on each of the first slice images is recorded as the corresponding synthesis center region; and based on a preset comparison region selection rule, an image region outside the synthesis center region and with the same shape as the synthesis center region is selected on each of the first slice images as the corresponding non-center comparison region; and each of the first slice images marked with the synthesis center region and the non-center comparison region is recorded as the corresponding second slice image; and all the second slice images corresponding to the current substance are combined to form a corresponding region identification data. The alignment region selection rule by default requires that the cell types contained in the non-center alignment region be consistent with those in the synthesis center region, and requires that the non-center alignment region does not contain any of the high-activity point coordinates x in the corresponding spatial region in the d-dimensional space. high Furthermore, the shape of the non-center alignment region must be consistent with that of the synthesis center region.

14. The method for analyzing plant metabolite synthesis centers and key genes according to claim 13, characterized in that, The method of identifying the corresponding key gene set by comparing the gene sequencing results of the synthesis centers and comparison regions of various observed substances specifically includes: Step 141: Take each type of observed substance in the observed substance list as the corresponding current substance; Step 142: Use the region identification data of the current substance as the current identification data; extract multiple second slice images marked with the synthesis center region and the non-center comparison region from the current identification data; and use the strongest active subset corresponding to the current substance as the corresponding current subset; The second slice image corresponds one-to-one with the sampled slice of the current subset; Step 143: Take each of the second slice images as the corresponding current slice image; use a preset laser cutting microscope to align the edges of the current slice image and the corresponding sampling slice with the object; and after alignment, cut the current sampling slice into corresponding regional samples according to the synthetic center region and the non-center comparison region marked on the current slice image to obtain a set of corresponding synthetic center region samples and non-center comparison region samples to form a corresponding first sample group. Step 144: Based on high-throughput sequencing technology, transcriptome sequencing is performed on the synthetic central region samples and the non-central alignment region samples of each of the first sample groups to obtain corresponding central region sequencing data and alignment region sequencing data; differential gene expression analysis is performed on the central region sequencing data and the alignment region sequencing data to obtain a set of differentially expressed genes with an upregulated gene reading in the central region sequencing data; GO enrichment analysis is performed on the differentially expressed gene set to screen the transcriptional regulatory factors of the current substance to obtain a corresponding set of first transcriptional regulatory factors; and KEGG pathway enrichment is performed on the differentially expressed gene set to screen the metabolic enzyme genes of the synthetic transformation pathway of the current substance to obtain a corresponding set of first metabolic enzyme genes. Step 145: Take the intersection of the first transcription regulatory factor sets of all the first sample groups to obtain the corresponding second transcription regulatory factor sets; and take the intersection of the first metabolic enzyme gene sets of all the first sample groups to obtain the corresponding second metabolic enzyme gene sets; and the second transcription regulatory factor sets and the second metabolic enzyme gene sets constitute the corresponding key gene sets.

15. An apparatus for performing the analytical method for plant metabolite synthesis centers and key genes according to any one of claims 1-14, characterized in that, The device includes: an analysis tool preparation module, an experimental material preparation module, an observation and sampling module, an adaptive training module, a simulation analysis module, and a report generation module; The analysis tool preparation module is used to construct partial differential equations to describe how the state of matter changes based on the spatiotemporal changes of the state of matter transportation and the state of matter synthesis or consumption, and to denote the current partial differential equation as the first equation; to design corresponding neural network models for the velocity field spatiotemporal function and the net generation rate spatiotemporal function in the first equation, and to denote the corresponding first model and second model; to set two corresponding ordinary differential equations for the first equation, and to denote the corresponding second equation and third equation; and to customize the data format for the sampled data sequence loaded with experimental sampling data, and to design the corresponding loss function L. ALL Based on the sampled data sequence, the second equation, the third-party process, and the loss function L... ALL Design a model training process for training the first model and the second model; The experimental material preparation module is used to extract the synthesis site of the target substance from the plant material as the experimental material; and to select an experimental reagent to spray the experimental material; and to store the treated experimental material in a preset moisturizing device. The observation and sampling module is used to take the material preservation start time as the start observation time, and customize the sampling time axis based on the preset observation duration and sampling time interval; and from the start observation time, every sampling time interval, the experimental material is sampled once based on the preset frozen section sampling rules to obtain the corresponding sampled section set; and the current sampled section set is divided into a first section subset and a second section subset and placed in a preset freezing device for storage; The adaptive training module is used to construct an observation substance list based on all upstream and downstream metabolites along the target substance synthesis pathway after sampling; extract sampling data from each type of observation substance in the observation substance list based on all the first slice subsets to obtain the corresponding sampling data sequence; and perform targeted training on the model parameters of the two models based on the sampling data sequence of each type of observation substance and the model training process to obtain the corresponding first version of the parameter set. The simulation analysis module simulates the spatiotemporal change state of the current material based on the first set of parameters for each type of observed material; and identifies the synthetic center region and non-center comparison region of the second slice subset of each type of observed material based on the simulation data; and identifies the corresponding key gene set by comparing the gene sequencing results of the synthetic center and the comparison region of each type of observed material. The report generation module constructs corresponding substance analysis reports based on the synthesis centers and key gene analysis information of various observed substances; and the metabolite analysis report corresponding to the target substance is composed of all the substance analysis reports and saved.

16. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-14; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-14.

Images