Method, device and equipment for constructing interaction network of stacking fermentation microorganisms and medium

Abstract

The invention provides an interaction network construction method and device for stacking fermentation microorganisms, equipment and a medium, and the method comprises the steps: constructing a quasi-dynamic differential equation, and decoupling the quasi-dynamic differential equation into a plurality of independent regression equations, the quasi-dynamic differential equation comprises a natural growth curve of microorganisms and a dependent growth curve of the influence of other microorganisms on the quasi-dynamic differential equation; solving the regression equation by using an add group lasso algorithm to obtain an independent expression quantity of each microorganism and at least one dependency expression quantity between the microorganisms and other microorganisms, and screening out significant variables in the microorganisms; and constructing a dynamic interaction network according to the independent expression quantity of the significant variable and the at least one dependency expression quantity. By implementing the method for constructing the interaction network of the stacking fermentation microorganisms, the problem that the interaction relationship among the stacking fermentation microorganisms is difficult to effectively analyze by a traditional method can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of microbiome data processing, in particular to a method and device for constructing an interaction network of microorganisms in a fermentation pile. BACKGROUND

[0002] Fermentation in a fermentation pile is a key step in the production of Maotai-flavor liquor, which determines the yield and quality of the fermented grains in the fermentation pile. The process is complex and shows significant heterogeneity. Microorganisms do not work independently in the fermentation pile, but interact through a complex network environment, which forms differences in community metabolic functions and directly affects the physicochemical indicators of the fermented grains. Therefore, revealing the interaction between microorganisms is an important foundation for studying fermentation in a fermentation pile and improving the yield and quality of the fermented grains.

[0003] However, traditional methods cannot capture the complex dynamic relationships and individualized characteristics at the gene level, and therefore cannot effectively analyze the complex topological structure contained in the interaction between microorganisms in a fermentation pile. How to reveal the micro-ecological interaction network in a fermentation pile has become a problem to be solved. SUMMARY

[0004] Based on this, the present application provides a method and device for constructing an interaction network of microorganisms in a fermentation pile, which can improve the problem that traditional methods cannot effectively analyze the interaction between microorganisms in a fermentation pile.

[0005] In a first aspect, the present application provides a method for constructing an interaction network of microorganisms in a fermentation pile, which comprises: constructing a pseudo-dynamic differential equation and decoupling it into a plurality of independent regression equations, wherein the pseudo-dynamic differential equation comprises a natural growth curve of a microorganism and a dependent growth curve of the microorganism affected by other microorganisms; using an adptive group lasso algorithm to solve the regression equations to obtain an independent expression of each microorganism and at least one dependent expression between the microorganism and other microorganisms, and screening out significant variables in the microorganisms; and constructing a dynamic interaction network according to the independent expression and at least one dependent expression of the significant variables.

[0006] In the first implementation manner of the first aspect, the step of constructing the dynamic interaction network according to the independent expression amount of the significant variable and the dependent expression amount comprises: determining the overall growth type of the significant variable in different samples, and the influence type and the influence strength of other microorganisms on the significant variable according to the independent expression amount and the dependent expression amount of the significant variable in different samples, wherein the overall growth type comprises rising and falling, and the influence type comprises inhibition and promotion; using nodes and edges to represent the interaction relationship between the significant variables and other significant variables respectively, and obtaining the dynamic interaction network between the significant variables by assigning different icon styles to distinguish the significant variables with different overall growth types, and assigning different icon styles to distinguish the interaction relationship with different influence types and influence strengths.

[0007] In the second implementation manner of the first aspect, the step of constructing the quasi-dynamic differential equation comprises: obtaining the abundance level of each microorganism in the heap fermentation in different samples through a plurality of samples; establishing a natural growth curve of the independent expression amount of the microorganism changing with the abundance level of the microorganism, and a dependent growth curve of the dependent expression amount of the microorganism changing with the abundance level of other microorganisms, to obtain the quasi-dynamic differential equation comprising the natural growth curve and the dependent growth curve.

[0008] In the third implementation manner of the first aspect, the step of decoupling the quasi-dynamic differential equation into a plurality of independent regression equations comprises: expanding the quasi-dynamic differential equation by using a truncated Legendre polynomial as a base function, wherein the truncated Legendre polynomial is used to describe the nonlinear interaction effect between the microorganisms; and decoupling to obtain a plurality of independent regression equations by estimating the variables and parameters of the expanded quasi-dynamic differential equation, wherein each regression equation can be used to independently calculate the expression amount of a single microorganism.

[0009] In the fourth implementation manner of the first aspect, the quasi-dynamic differential equation comprises:

[0010]

[0011] wherein i = 1, …, m; j = 1, …, n, m represents the total number of samples, and n represents the total number of microbial OTUs, is the natural growth curve of the OTU, is the allometric growth curve of the OTUj, is the allometric growth curve of the OTUj in the ith sample, g ji is the abundance level of the OTUj in the ith sample, E i is the abundance level of all microbial OTUs in the ith sample, and j' is an OTU other than the OTUj, wherein, yj'is the growth curve of the j'th OTU depending on the j'th OTU, is the allometric growth curve of the j'th OTU in the i'th sample, is the abundance level of the j'th OTU in the i'th sample, d represents the number of OTUs that have an effect on the j'th OTU, Θ j and Θ j←j' are equation parameters.

[0012] In combination with the first aspect, in a fifth implementation manner of the first aspect, after constructing the dynamic interaction network according to the independent expression amount of the significant variable and the at least one dependent expression amount, the method further comprises: calculating the eigenvector centrality of each node in the interaction network; measuring the influence of the significant variable corresponding to each node by using the eigenvector centrality; and determining the significant variable with the largest influence in the interaction network as the core microorganism.

[0013] In combination with the fifth implementation manner of the first aspect, in a sixth implementation manner of the first aspect, the mathematical expression of the step of calculating the eigenvector centrality of each node in the interaction network comprises:

[0014]

[0015] wherein, x v is the eigenvector centrality of the node v, λ is the maximum eigenvalue of the adjacency matrix of the interaction network, t e M(v) is all nodes adjacent to the node v in the interaction network, x t is the eigenvector centrality of the node t, V is all nodes of the interaction network, a v,t is used to represent whether the nodes v and t are adjacent, if the node v is adjacent to the node t, a v,t = 1, if the node v is not adjacent to the node t, a v,t = 0.

[0016] In a second aspect, the present application provides an interaction network construction device, which comprises: an equation construction unit, configured to construct a pseudo-dynamic differential equation and decouple the pseudo-dynamic differential equation into a plurality of independent regression equations, wherein the pseudo-dynamic differential equation comprises a natural growth curve of a microorganism and a dependent growth curve of the microorganism depending on other microorganisms; an equation decoupling unit, configured to solve the regression equations by using an adptive group lasso algorithm to obtain an independent expression amount of each microorganism and at least one dependent expression amount between the microorganism and other microorganisms, and screen out significant variables in the microorganisms; and a network construction unit, configured to construct a dynamic interaction network according to the independent expression amount of the significant variable and the at least one dependent expression amount.

[0017] In a first implementation form of the second aspect, the network construction unit is configured to: determine, according to the independent expression and the dependent expression of the significant variable in different samples, a global growth type of the significant variable in different samples, and an influence type and an influence strength of other microorganisms on the significant variable, wherein the global growth type includes an increase and a decrease, and the influence type includes inhibition and promotion; and obtain a dynamic interaction network between the significant variables by using nodes and edges to represent interaction relationships between the significant variables and other significant variables, and by assigning different icon styles to distinguish the significant variables of different global growth types, and assigning different icon styles to distinguish the interaction relationships of different influence types and influence strengths.

[0018] In a second implementation form of the second aspect, the equation construction unit is configured to: obtain, by using a plurality of samples, abundance levels of each microorganism in different samples in the cumulative fermentation; establish a natural growth curve of the independent expression of the microorganism with respect to the abundance level of the microorganism, and a dependent growth curve of the dependent expression of the microorganism with respect to the abundance level of other microorganisms, to obtain a quasi-dynamic differential equation including the natural growth curve and the dependent growth curve.

[0019] In a third implementation form of the second aspect, the equation construction unit is configured to: expand the quasi-dynamic differential equation by using a truncated Legendre polynomial as a base function, wherein the truncated Legendre polynomial is used to describe a nonlinear interaction effect between the microorganisms; and decouple, by estimating variables and parameters of the expanded quasi-dynamic differential equation, a plurality of independent regression equations, wherein each regression equation can be used to independently calculate the expression of a single microorganism.

[0020] In a fourth implementation form of the second aspect, the quasi-dynamic differential equation is:

[0021]

[0022] wherein i = 1, …, m; j = 1, …, n, m represents a total number of samples, and n represents a total number of microorganism OTUs, is a natural growth curve of the OTU, is a growth curve of the OTUj, is a growth curve of the OTUj in the i-th sample, g ji is an abundance level of the OTUj in the i-th sample, E i is an abundance level of all microorganism OTUs in the i-th sample, j' is an OTU other than the OTUj, is a dependent growth curve, representing an influence of the j'-th OTU on the j-th OTU, is a growth curve of the OTUj' (j'≠j) in the i-th sample, is the abundance level of OTUj' (j'≠j) in the i-th sample, d represents the number of OTUs that have an influence on OTUj, Θ j and Θ j←j' are equation parameters.

[0023] With reference to the second aspect, in a fifth implementation form of the second aspect, the network constructing unit is further configured to: calculate eigenvector centrality of each node in the interaction network; measure influence of the significant variable corresponding to each node by using the eigenvector centrality; and determine the significant variable with the largest influence in the interaction network as the core microorganism.

[0024] With reference to the fifth implementation form of the second aspect, in a sixth implementation form of the second aspect, a mathematical expression of the step of calculating the eigenvector centrality of each node in the interaction network comprises:

[0025]

[0026] wherein x v is the eigenvector centrality of node v, λ is the largest eigenvalue of the adjacency matrix of the interaction network, t∈M(v) is all nodes adjacent to node v in the interaction network, x t is the eigenvector centrality of node t, V is all nodes of the interaction network, a v,t is used to represent whether nodes v and t are adjacent, a v,t = 1 if nodes v and t are adjacent, and a v,t = 0 if nodes v and t are not adjacent.

[0027] In a third aspect, the present application further provides an interaction network construction device, the interaction network construction device comprising a processor and a memory connected through a bus; the processor is configured to execute a plurality of instructions; and the memory is configured to store the plurality of instructions, the instructions being adapted to be loaded and executed by the processor to perform the interaction network construction method according to the first aspect or any one of the implementation forms of the first aspect.

[0028] In a fourth aspect, the present application further provides a computer readable storage medium, the computer readable storage medium storing a plurality of instructions, the instructions being adapted to be loaded and executed by a processor to perform the interaction network construction method according to the first aspect or any one of the implementation forms of the first aspect.

[0029] In summary, the application provides a method, device, equipment and medium for constructing an interaction network of stacked fermentation microorganisms. The application captures the complex dynamic relationship and individualized characteristics among microorganisms by constructing a quasi-dynamic differential equation including a natural growth curve of stacked fermentation microorganisms and a dependent growth curve of the influence of other microorganisms thereon, and then decouples the quasi-dynamic differential equation into a plurality of independent regression equations, so that each regression equation can be used to independently calculate the expression amount of a single microorganism. The adaptive group lasso algorithm is used to solve the regression equation to obtain the independent expression amount of each microorganism and at least one dependent expression amount between the microorganism and other microorganisms, and significant variables in the microorganism are screened out. Based on the independent expression amount and at least one dependent expression amount of the significant variables, a dynamic interaction network between the stacked fermentation microorganisms is constructed, so as to effectively characterize the dynamic interaction relationship between the stacked fermentation microorganisms, and provide a reasonable reference for revealing core microorganisms and regulating an open stacked fermentation process. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 A flowchart of the method for constructing an interaction network in an embodiment of the application;

[0031] Figure 2 A structural diagram of an interaction network in an embodiment of the application;

[0032] Figure 3 An effect curve decomposition diagram of a quasi-dynamic differential equation in an embodiment of the application;

[0033] Figure 4 An effect curve decomposition diagram between bacterial modules in round A in an embodiment of the application;

[0034] Figure 5 An effect curve decomposition diagram between bacterial modules in round B in an embodiment of the application;

[0035] Figure 6 An effect curve decomposition diagram between bacterial modules in round C in an embodiment of the application;

[0036] Figure 7 An effect curve decomposition diagram between bacterial modules in round D in an embodiment of the application;

[0037] Figure 8 A module effect curve decomposition diagram in round A in an embodiment of the application;

[0038] Figure 9 A module effect curve decomposition diagram in round B in an embodiment of the application;

[0039] Figure 10A module effect curve disassembly diagram of C round in an embodiment of the present application;

[0040] Figure 11 A module effect curve disassembly diagram of D round in an embodiment of the present application;

[0041] Figure 12 An interaction network of bacterial modules of each round in an embodiment of the present application;

[0042] Figure 13 An interaction network of bacterial core modules of B round in an embodiment of the present application;

[0043] Figure 14 An interaction network of bacterial and fungal mixed modules of each round in an embodiment of the present application;

[0044] Figure 15 An interaction network of M7 in the core module of B round in an embodiment of the present application;

[0045] Figure 16 A correlation network of M7 in the core module of B round in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0047] At present, the traditional method is difficult to capture the complex dynamic relationship between genes and its personalized characteristics, so as to effectively analyze the complex topological structure of the interaction relationship between the microorganisms in the heap fermentation, how to reveal the micro ecological interaction network of the heap fermentation becomes a problem to be solved.

[0048] In this regard, the present application constructs a pseudo-dynamic differential equation including a natural growth curve and a dependent growth curve, to describe the natural growth of the microorganisms in the heap fermentation and the dependent growth caused by the interaction between the microorganisms, which is supported by theories such as ecology and evolutionary game theory:

[0049] On the one hand, the natural growth curve and the dependent growth curve can be understood by the inherent growth and dependent growth in the LV model in ecology. In ecology, the LV model is often used to describe the dynamic interaction relationship between predators and prey in biological systems, reflecting the increase and decrease of the population size of the two. The LV model can be represented by the following system of differential equations:

[0050]

[0051] where x is the number of prey, y is the number of predators, a, b, g, d represent coefficients related to the interaction between the two species, a represents the rate of natural growth of prey, b represents the probability of encountering predators, g represents the natural death of predators, and d can be understood as the coefficient of the conversion of food into predators. The first equation is the change in the number of prey, and the final change in the number of prey is determined by the natural growth a x of its reproduction and the dependent growth b x y of being preyed upon. The second equation represents the change in the number of predator populations, and according to the above equation, the change in the size of the predator population is the growth d x y of the predator prey, minus the natural death of the predator y. It can be seen that according to the Lotka-Volterra equation, the change in the number of a species is determined by its own influence (intrinsic growth) and the influence of other species on it (dependent growth).

[0052] On the other hand, the natural growth curve and the dependent growth curve can also be understood by the independent component and the dependent component of evolutionary game theory. Game theory combined with evolutionary theory forms evolutionary game theory, in which the process of reaching equilibrium Nash equilibrium is defined by evolutionarily stable strategy, which allows the value of a variable to be decomposed into two components, the first component is the value of the variable due to its own strategy (called independent component), and the second component is the value of the variable due to the influence of other variables on it (called dependent component). Assume that each sample of the stacked fermentation forms an ecological system, and the interaction between the microorganisms in the system is considered to follow game theory, that is, each microorganism is like a player in a game of game, always trying to get the maximum benefit (variable value / expression level) in the interaction with other players. This requires each player to adjust and adopt its own optimal strategy based on the strategy of the other players, and this process continues until the system reaches Nash equilibrium.

[0053] For example, assume that the abundance data of the microorganisms in the stacked fermentation includes the abundance levels of n OTUs in m samples after sequencing, which can be represented as an n x m matrix, and the quasi-dynamic differential equation qdODEs constructed based on the abundance data can have the following form:

[0054]

[0055] where i = 1, …, m; j = 1, …, n, m represents the total number of samples, and n represents the total number of microbial OTUs, is the natural growth curve of the OUT, is the allometric growth curve of OTUj, is the growth curve of OTUj in the ith sample, g ji is the abundance level of OTUj in the ith sample, E i is the abundance level of all microbial OTUs in the ith sample, j' is an OTU other than OTUj, is the dependence growth curve, representing the effect of the j'th OTU on the jth OTU, is the growth curve of OTUj'(j'≠j) in the ith sample, is the abundance level of OTUj'(j'≠j) in the ith sample, d represents the number of OTUs that have an effect on OTUj, Θ j and Θ j←j' are equation parameters, Θ j and Θ j←j' can be obtained by parameter estimation. Before parameter estimation, Q j (·) is a function containing unknown parameters Θ j , corresponding to the intrinsic growth part in the Lotka-Volterra equation, Q j←j' (·) is a function containing unknown parameters Θ j←j' , corresponding to the dependent growth part in the Lotka-Volterra equation.

[0056] In comparison with the general ODE model, its form can be expressed as follows:

[0057]

[0058] where t∈[t0, T])0≤t0<T<∞), t represents time, X(t) = (X1(t), …, X p (t)) T is the vector of the abundance levels of OTU1, …, n at time t, and X'(t) is the first derivative of X(t). The function Q serves as a connection function, quantifying the regulatory effect of the regulatory gene on the expression change of the target gene, which depends on the parameter vector θ.

[0059] The qdODEs of the present application decouple the quasi-dynamic differential equations into multiple independent regression equations, and the decoupled property they embody allows the multi-dimensional ODE to be split into independent one-dimensional regression problems, facilitating parameter estimation and variable selection during equation construction, greatly improving the efficiency of equation construction and subsequent establishment of interaction networks.

[0060] It should be noted that the interaction network construction method provided in the present application can be applied to an interaction network construction device or an interaction network construction apparatus. The interaction network construction device or the interaction network construction apparatus can be a server, a computer, a terminal device, a controller, a processor, or the like, which can implement the interaction network construction method of the present application. In addition, the interaction network construction device or the interaction network construction apparatus can interact with other servers, terminal devices, controllers, processors, or the like, and execute the interaction network construction method provided in the present application.

[0061] Reference Figure 1 The present application also provides an embodiment of a method for constructing an interaction network of stacked fermentation microorganisms. The embodiment takes an interaction network construction apparatus as an execution subject, and specifically includes the following steps:

[0062] 110: constructing a pseudo-dynamic differential equation and decoupling it into a plurality of independent regression equations, wherein the pseudo-dynamic differential equation includes a natural growth curve of a microorganism and a dependent growth curve of the microorganism affected by other microorganisms;

[0063] 120: using an adptivegrouplasso algorithm to solve the regression equations to obtain independent expression amounts of each microorganism and at least one dependent expression amount between the microorganisms, and screening out significant variables in the microorganisms;

[0064] 130: constructing a dynamic interaction network according to the independent expression amounts and the at least one dependent expression amount of the significant variables.

[0065] In the step of constructing the interaction network, one node is used to represent one microorganism, and one edge is used to represent the interaction relationship between the microorganisms. Since the independent expression amount and the dependent expression amount are characteristics of the microorganism affected by itself and other microorganisms, respectively, the growth of each microorganism and the interaction relationship between the microorganisms can be determined according to the positive or negative situation and the numerical value of the independent expression amount and the at least one dependent expression amount of the microorganism in different samples, so as to distinguish different microorganisms and the interaction relationship between the microorganisms through different forms, so as to construct an interaction network that dynamically reflects the growth of each microorganism and the interaction relationship between the microorganisms.

[0066] In an implementation manner, the interaction network generated through the step 130 can be an idopNetwork (Inference of Dynamic Regulatory Networks). The idopNetwork is a holographic, dynamic, all-around, individualized, and pseudo-dynamic interaction network. Holographic means that the interaction between variables can be bidirectional, signed, and directional. For reference Figure 2Dynamic means that the interaction network is allowed to change nonlinearly over time and space, all-around means that the potential interaction relationship between each variable and other variables can be found, individualized means that each observation sample has a specific interaction network, quasi-dynamic processing means that due to the difficulty of obtaining time series data, the real-time data is converted into quasi-dynamic data by introducing the scaling law, and the functional relationship between each variable and the niche index is obtained by integration.

[0067] The step of generating the interaction network by step 130 comprises:

[0068] 131: determining the overall growth type of the significant variable in different samples, and the influence type and influence strength of other microorganisms on it according to the independent expression amount and dependent expression amount of the significant variable in different samples;

[0069] 132: obtaining the dynamic interaction network between the significant variables by using nodes and edges to represent the interaction relationship between the significant variables and other significant variables, and by assigning different icon styles to distinguish the significant variables of different overall growth types, and by assigning different icon styles to distinguish the interaction relationship of different influence types and influence strengths.

[0070] In step 131, the sum of the independent expression amount and the dependent expression amount of the significant variable is calculated, and the overall growth type of each significant variable is determined according to the positive or negative of the sum, the overall growth type includes rising and falling, if the sum is positive, it indicates that the overall growth type of the microorganism corresponding to the significant variable is rising, if the sum is negative, it indicates that the overall growth type is falling; the influence type of other significant variables on the significant variable is determined according to the positive or negative of the dependent expression amount of the significant variable, the influence type includes inhibition and promotion, if a dependent expression amount is positive, it indicates that the influence type of a certain other significant variable on the significant variable is promotion, if a dependent expression amount is negative, it indicates that the influence type of a certain other significant variable on the significant variable is inhibition; the influence strength of other significant variables on the significant variable is determined according to the numerical value of the dependent expression amount of the significant variable in different samples, the larger the numerical value, the stronger the influence, the smaller the numerical value, the weaker the influence;

[0071] In step 132, if the overall growth type of the significant variable is rising, the node of the significant variable is represented by red color, if the overall growth type of the significant variable is falling, the node of the significant variable is represented by blue color; the other significant variables corresponding to each dependent expression amount are determined, the nodes corresponding to the other significant variables are taken as the starting points to the directed edge of the node corresponding to the significant variable, that is, an arrow from the other significant variable to the significant variable is established between the significant variable and the other significant variable, indicating that the other significant variable has an influence relationship on the significant variable; if the influence type of the other significant variable on the significant variable is promotion, the edge between the significant variables is represented by red color, if the influence type of the other significant variable on the significant variable is inhibition, the edge between the significant variables is represented by blue color; the interaction relationship of different influence strengths is represented by edges of different thicknesses, if the dependent expression amount corresponding to a certain other significant variable is larger, the edge between the significant variables is thicker, if the dependent expression amount corresponding to a certain other significant variable is smaller, the edge between the significant variables is thinner.

[0072] For step 110, the application further provides a specific implementable manner, comprising:

[0073] 111: obtaining the abundance level of each microorganism in different samples through the accumulation fermentation of a plurality of samples;

[0074] 112: establishing a natural growth curve of the independent expression amount of the microorganism changing with the abundance level of the microorganism and a dependent growth curve of the dependent expression amount of the microorganism changing with the abundance level of the other microorganism, to obtain a quasi-dynamic differential equation including the natural growth curve and the dependent growth curve;

[0075] 113: expanding the quasi-dynamic differential equation by using a truncated Legendre polynomial as a base function, wherein the truncated Legendre polynomial is used to describe the nonlinear interaction effect between the microorganisms;

[0076] 114: decoupling to obtain a plurality of independent regression equations by estimating the variables and parameters of the expanded quasi-dynamic differential equation, wherein each regression equation can be used to independently calculate the expression amount of a single microorganism.

[0077] Wherein, the abundance level of the microorganism in different samples is used to fit the and Then, the and represent the natural growth curve and the dependent growth curve The natural growth curve is used to represent the curve of the independent expression amount of the microorganism changing with the abundance level of the microorganism, and the dependent growth curve The dependence expression of the microorganism changes with the abundance level of other microorganisms. In order to express the specific Q function, i.e. the natural growth curve and the dependence growth curve, the present application applies the truncated series expansion and the base function φ(x) to approximate the function Q in the above formula, wherein q i (x) is the Q function of the ith gene, which can be expressed as follows:

[0078]

[0079] The foregoing base function can be selected from a smooth spline function, a regression spline function, a penalized spline function, a local polynomial, and a Legendre polynomial, etc. The selection of the base function is very important, and the Legendre polynomial is preferably selected as the smooth function, which has the advantages of function orthogonality, applicability to fitting the data sparsity caused by the time-space heterogeneity and the ecological niche heterogeneity in the accumulation fermentation process, ability to estimate a complex high-order curve, good convergence, and high calculation efficiency, etc. If the r-order Legendre polynomial is selected, the unknown parameter Θ j and the unknown parameter Θ j←j' are the coefficients of the r-order Legendre polynomial, and the Q function can be expressed as follows:

[0080]

[0081] The Legendre polynomial is defined in the interval [-1, 1], and in the above polynomial, the independent variable x can represent the time variable t, and the actual measurement time needs to be corrected to be between [-1, 1], and the correction process is as follows,

[0082]

[0083] wherein t min and t max are the first and last time points in the time series, respectively.

[0084] For step 120, due to the sparsity characteristics of the interaction network and the large amount of OTU data characteristics caused by the niche heterogeneity of the microorganisms in the accumulation fermentation, the interaction network needs to be reduced to extract valuable interaction information in the network.

[0085] For this purpose, the present application provides a specific implementable manner of step 120, which determines the genes with significant interactions in the pseudo-dynamic differential equation by introducing the high-dimensional variable selection technology Adaptive Group LASSO. Let and generate n groups of parameters, and the groups with parameters not equal to 0 are screened out by variable selection, and the parameters of the groups are denoted as β ji ,i,j=1,...,n.

[0086] The aforementioned step of selecting variables of the quasi-dynamic differential equation to retain microorganisms with significant interaction relationship comprises: performing preliminary screening on the variables in the quasi-dynamic differential equation based on Group LASSO; and further screening significant variables with significant interaction relationship from the variables obtained through the preliminary screening based on Adaptive Group LASSO, so as to retain the significant variables with significant interaction relationship in the microorganisms.

[0087] In the application, Adaptive Group LASSO is used for screening, which is a more flexible penalty strategy, improves the convergence rate of parameter estimation, and more accurately identifies significant variables.

[0088] The Group LASSO estimator can be obtained by minimizing the following penalty-weighted least squares criterion

[0089]

[0090] wherein λ i is a penalty parameter, which can be determined using the BIC criterion; and ||·||2 is the Euclidean norm. The optimization criterion is similar to LASSO at the group level: depending on the penalty parameter λ i , the coefficients of a group of prediction variables can be 0 and thus cannot be added to the model; when the number of groups is 1, the Group LASSO will become LASSO.

[0091] Based on the Group LASSO result, adaptive group LASSO is used to further screen the main variables, and the weight w ji of the penalty coefficient can be obtained by the following method:

[0092] If , w ji = 0.

[0093] The adaptive group LASSO estimator can be obtained by minimizing the following penalty-weighted least squares criterion

[0094]

[0095] wherein λ i is a penalty parameter, which can be determined using the BIC criterion; and Z i is a boundary constraint weight function in the group LASSO for improving the convergence rate of parameter estimation. The genes screened by the adaptive group LASSO are used as candidate genes for subsequent construction of the interaction network.

[0096] The nodes and edges of a network form a set G = (V, E), where V is the set of vertices and E is the set of edges, consisting of pairs of elements from V. Taking a network containing 3 species as an example, the set of nodes and edges for this network can be written in the following form:

[0097] V = {x, y, z}

[0098] E={(x,y),(x,z),(y,x),(y,z),(z,x),(z,y)}

[0099] Here, the network represented by this set is a directed graph, and there is a distinction between (x,y) and (y,x), which respectively represent the influence of species x on species y and the influence of species y on species x. That is, they indicate different directions, and the edges in this case are called arcs.

[0100] Therefore, the interaction network based on qdODEs can be rewritten in the following form:

[0101] V = {g1, g2, ..., g} m}

[0102] E = {(g1,g2),(g1,g3),…,(g...} m-1 ,g m ),(g m ,g1)}

[0103] Among them, the quasi-dynamic differential equation qdODEs V represents a node in the network; E represents the influence of the j'-th OTU on the j-th OTU, and this arc can be represented as (g j' ,g j ).

[0104] In an implementable manner, due to the complex interaction relationship between the fermentation microorganisms, the interaction network of the fermentation microorganisms is more complex and larger than the interaction network of general species, and with the increase of the complexity and size of the interaction network, the key nodes in the network play a very important role in the in-depth understanding of the complex network system. In this regard, the eigenvector centrality is used to measure the node influence in the application to mine the core microorganisms in the interaction network, because the eigenvector centrality measures the relative importance of the node in the network, and its core principle is that the importance of a node is determined by the importance of its neighbors. Specifically, in the calculation of the index, the contribution of the connection to the high centrality neighbor to the promotion of the centrality of the node itself is much greater than the connection to the low centrality neighbor, so the high eigenvector centrality score finally calculated means that the node is connected to many nodes with high centrality scores. Unlike the degree centrality and other indexes that only focus on local connection, the eigenvector centrality emphasizes that the value of the node is recursively defined by the value of its high-quality neighbors, reflecting the global influence of the node in the entire network structure. The steps of determining the core microorganism include:

[0105] After constructing the dynamic interaction network according to the independent expression amount of the significant variable and the dependent expression amount, the method further includes: calculating the eigenvector centrality of each node in the interaction network; using the eigenvector centrality to measure the influence of the significant variable corresponding to each node; determining the significant variable with the largest influence in the interaction network as the core microorganism. Further, only the other target microorganisms that have an interaction relationship with the core microorganism can be retained, and an interaction network with the core microorganism as the hub can be constructed according to the interaction relationship between the core microorganism and the other target microorganisms.

[0106] In addition, different icon styles can be used to distinguish the microorganisms and non-core microorganisms in the interaction network, wherein the mathematical expression of the step of calculating the eigenvector centrality of each node in the interaction network includes:

[0107]

[0108] Wherein, x v is the eigenvector centrality of node v, λ is the maximum eigenvalue of the adjacency matrix of the interaction network, t ∈ M(v) is all nodes adjacent to node v in the interaction network, M(v) is the neighbor set of v, x t is the eigenvector centrality of node t, V is all nodes of the interaction network, a v,t is used to represent whether nodes v and t are adjacent, if nodes v and t are adjacent, a v,t = 1, if nodes v and t are not adjacent, a v,t = 0.

[0109] In an implementable manner, based on the quasi-dynamic differential equation, according to the allometric growth curve of the microorganism and the size relationship with the natural growth curve , two expression modes of the microorganism can be determined, including the intrinsic growth type and the recession type. If the expression mode of the microorganism is determined as the intrinsic growth type, if the expression mode of the microorganism is determined as the recession type.

[0110] Further, according to the size relationship between the dependent growth curve the allometric growth curve and the natural growth curve , four expression modes of the microorganism can be determined, including: the actual growth but intrinsic recession type (A), the actual growth and intrinsic growth type (B), the actual recession and intrinsic recession type (C), and the actual recession but intrinsic growth type (D).

[0111] The aforementioned four expression modes can be decomposed with reference to the effect curve diagram of Figure 3 , in which the red line represents the natural growth curve , the green line represents the dependent growth curve , and the blue line represents the allometric growth curve , and it should be noted that in this paper, the effect curve diagram of all modules, sub-modules or genes is represented by red color for the self-effect of the module, sub-module or gene, green color for the influence of other modules, sub-modules or genes, and blue color for the actual observation value of the module, sub-module or gene, which will not be described hereinafter.

[0112] In an implementable manner, based on the quasi-dynamic differential equation of each microorganism, the dependent expression amount generated by the mutual influence between the microorganisms can be determined, assuming that OTU a and OTU b influence each other, and taking the mean value of the overall effect, the following Table 1 shows the relationship between OTU a and OTU b:

[0113]

[0114] Table 1 describes the interaction mode between OTU a and OTU b

[0115] Based on the aforementioned quasi-dynamic differential equation, the complex dynamic relationship between genes can be better captured as a whole by analyzing the multiple effect curve decomposition diagrams of the microorganisms in each round, which helps to more effectively analyze the interaction relationship between the microorganisms in the heap fermentation. Next, based on the quasi-dynamic differential equation, the effect curve decomposition diagrams between the bacterial modules in each round and the effect curve decomposition diagrams between the bacterial and fungal modules in each round are analyzed, in particular:

[0116] refer to Figures 4 to 7 The decomposition plot of the effect curves between bacterial modules in rounds A to D shows that, overall, the intrinsic growth effect (red line) of each module is greater than its observed effect (blue line). However, in round C, more instances were found where the intrinsic growth effect (red line) was less than its observed effect (blue line), meaning that round C produced a greater promoting effect among bacterial modules compared to other rounds (green line). A similar situation existed in round A, but the promoting effect was even lower.

[0117] refer to Figures 8 to 11 The effect curves between the bacterial and fungal modules in rounds A to D are decomposed. Overall, the intrinsic growth effect (red line) of each module is greater than its observed effect (blue line). However, in each round, some modules showed more cases where the intrinsic growth effect (red line) was less than its observed effect (blue line), indicating that they may have gained an advantage in the stacking fermentation process.

[0118] In one possible implementation, network properties of the interacting network are extracted. These properties help to understand the structure and function of the network and are very important for analyzing complex network systems such as social networks, computer networks, and biological networks. They can be quickly calculated using functions in the r package igraph.

[0119] Network attributes are metrics that describe the structure and characteristics of a network. Network attributes include at least one of the following: node degree, degree distribution, average path length, eigenvector centrality, diameter, clustering coefficient, connectivity, community structure, and centrality.

[0120] For example, node degree refers to the number of edges a node is connected to. In an undirected network, node degree equals the number of edges connected to that node; in a directed network, node degree is divided into in-degree and out-degree, representing the number of edges connected to and originating from that node, respectively. Degree distribution describes the distribution of node degrees in a network, representing the probability that a randomly selected node in the network has a degree of k. Different network models (such as random networks and scale-free networks) have different degree distributions. Average path length refers to the average of the shortest path lengths between any two nodes in a network, reflecting the overall connectivity and information propagation efficiency of the network. Generally, networks with shorter average path lengths transmit information faster.

[0121] In combination with the aforementioned quasi-dynamic differential equation, interaction network, core microorganism, and network attribute, the application further provides some practical data analysis results to better understand the foregoing:

[0122] The network attribute between the bacterial modules in each round is shown in Table 2, for example:

[0123]

[0124] Table 2: Network attribute between bacterial modules in each round

[0125] Interaction network between bacterial modules in each round is shown in the following reference: Figure 12 In each round, the left graph is the interaction network (red nodes represent that the overall growth type of the module is rising, and blue nodes represent that the overall growth type of the module is falling), the upper right graph is the characteristic vector centrality score of each module, and the lower right graph is the independent expression and dependent expression of each module;

[0126] In each round in the network shown in Figure 12 The upper right graph on the right shows that the characteristic vector centrality score of each module decreases from left to right, and the greater the characteristic vector centrality score of the module, the greater the global influence of the module in the entire network structure. Specifically, in round A, module M6, module M3, and module M4 are the top three core modules; in round B, module M7, module M8, and module M5 are the top three core modules; in round C, module M1, module M7, and module M4 are the top three core modules; and in round D, module M7, module M5, and module M6 are the top three core modules. From the above, it can be seen that the core module in each round is different, and the core module in each round is not the same. Figure 12 It can also be seen that the sparsity of the network between the modules in each round is different, the network between the modules in round C is the most sparse, the network in round D is the least sparse, indicating that the regulation relationship between the network modules decreases from round A to round C, and the regulation between the network modules is more from round C to round D, because the abundance of some OTUs is 0 in a specific round, and the OTU does not have an interaction effect on any OTU, so the module corresponding to the OTU is removed when the network graph is drawn.

[0127] The network attribute in the bacterial module in each round is shown in Table 3, for example:

[0128]

[0129]

[0130]

[0131] Table 3: Network attribute in the bacterial module in each round

[0132] Interaction network in the bacterial module in each round is shown in the following reference: Figure 13, the left figure is the interaction network between each microorganism in the B round module M8 (red nodes represent the overall growth type of the microorganism is rising, and blue nodes represent the overall growth type of the microorganism is falling), and the right upper figure is the characteristic vector centrality score of each microorganism, and the right lower figure is the independent expression amount and dependent expression amount of each microorganism;

[0133] In Figure 13 , the right upper figure shows that the characteristic vector centrality score of each microorganism decreases from left to right, and the greater the characteristic vector centrality score of the microorganism, the greater the global influence of the microorganism in the entire network structure. Specifically, Bosea, Arthrobacter and Lactococcus have the highest characteristic vector centrality, and are core microorganisms. In this module, Bosea inhibits most of the bacteria in the module, and only a small part of the bacteria promotes other bacteria.

[0134] The network properties between the bacterial and fungal modules in each round are shown in Table Four

[0135]

[0136] Table Four Network Properties Between Bacterial and Fungal Modules in Each Round

[0137] The interaction network between bacterial and fungal modules in each round is shown in Table Four Figure 14 , the network between the bacterial and fungal modules in A round shows that M4 is the core module, followed by M5, M3 and M1; the network between the bacterial and fungal modules in B round shows that M7 is the core module, followed by M8, M4 and M6; the network between the bacterial and fungal modules in C round shows that M6 is the core module, followed by M4, M3 and M1; the network between the bacterial and fungal modules in D round shows that M8 is the core module, followed by M1, M4 and M7. The number of edges between modules in D round is the most, reaching 0.41, followed by A round, while the number of edges between modules in B and C rounds is less, indicating that there are more regulatory relationships between modules in rounds A and D. At the same time, the network distance between modules in D round is the shortest (1.206897), indicating that the network between modules in D round is relatively more complex and close.

[0138] The network properties within the bacterial and fungal modules in each round are shown in Table Five

[0139]

[0140]

[0141] Table Five Network Properties Within Bacterial and Fungal Modules in Each Round

[0142] The interaction network within the bacterial and fungal modules in each round is shown in Table FiveFigure 15 Take M7 in round B as an example, the left figure is the interaction network between each microorganism in the module M7 in round B, the upper right figure is the eigenvector centrality score of each microorganism, and the lower right figure is the proportion of each relationship type in the network (the red part is the proportion of the relationship between bacteria and bacteria, the green is the proportion of the relationship between bacteria and fungi, and the blue is the proportion of the relationship between fungi and fungi);

[0143] In Figure 15 , Cyberlindnera and Rhodococcus in the module M7 play the core microorganisms in fungi and bacteria respectively, and the network calculated by the Adaptive Group LASSO penalty highlights the interaction network relationship in the M7 module with the two microorganisms as the hub. In this relationship, the color of the line represents the positive (red) and negative (blue) influence, the arrow represents the direction of the influence, and the thickness of the line represents the strength of the influence.

[0144] Therefore, the network has all the information reflecting the microbial network in the open fermentation process, which plays a crucial reference role in the regulation of complex microbial solid-state fermentation networks. In the network, the core microorganisms Cyberlindnera and Rhodococcus inhibit and compete with each other, and dominate and inhibit part of other microorganisms in fungi and bacteria respectively, in addition, microorganisms such as Enterococcus and Brevibacterium are affected by the growth and decline of the two core microorganisms. In summary, if the microorganisms in this module need to be regulated, the fermentation process can be optimized according to the growth characteristics of the hub microorganisms.

[0145] For the network in each round of the bacteria and fungi module, the present application also provides a comparative example, referring to Figure 16 , the comparative example also takes the microorganisms in the module M7 in round B as an example, uses the same data to draw and analyze the correlation network commonly used for bacterial interaction in R studio using the igraph package (version 1.5.0.1) and the psych package (version 2.4.3), and classifies bacteria and fungi. It is not difficult to find that the correlation network only calculates the correlation between two microorganisms, lacks the direction of influence, and contains a large amount of redundant information caused by the lack of penalty calculation. In addition, due to the lack of analysis of different ecological niches of samples, a large number of positive correlations appear between microorganisms, which cannot reflect the real interaction relationship between microorganisms in the real fermentation process, and cannot screen the core microorganisms. Therefore, it cannot be applied to the interaction network analysis of the open solid-state fermentation process with extremely large ecological niche heterogeneity such as pile fermentation.

[0146] The application also provides an interaction network construction device, which comprises: an equation construction unit, configured to construct a pseudo-dynamic differential equation and decouple the pseudo-dynamic differential equation into a plurality of independent regression equations, wherein the pseudo-dynamic differential equation comprises a natural growth curve of a microorganism and a dependent growth curve of the microorganism affected by other microorganisms; an equation decoupling unit, configured to solve the regression equations by using an adptive group lasso algorithm to obtain independent expression amounts of each microorganism and at least one dependent expression amount between the microorganism and other microorganisms, and screen out significant variables in the microorganisms; and a network construction unit, configured to construct a dynamic interaction network according to the independent expression amounts and the at least one dependent expression amount of the significant variables.

[0147] In an implementation, the network construction unit is specifically configured to: determine overall growth types of the significant variables in different samples and influence types and influence strengths of other microorganisms on the significant variables according to the independent expression amounts and the dependent expression amounts of the significant variables in the different samples, wherein the overall growth types include rising and falling, and the influence types include inhibition and promotion; and obtain a dynamic interaction network between the significant variables by using nodes and edges to respectively represent interaction relationships between the significant variables and other significant variables, and by using different icon styles to distinguish the significant variables of different overall growth types and to distinguish the interaction relationships of different influence types and influence strengths.

[0148] In an implementation, the equation construction unit is specifically configured to: obtain abundance levels of each microorganism in different samples by a plurality of samples in a heap fermentation; establish a natural growth curve of an independent expression amount of a microorganism changing with an abundance level of the microorganism and a dependent growth curve of a dependent expression amount of the microorganism changing with an abundance level of other microorganisms, to obtain a pseudo-dynamic differential equation comprising the natural growth curve and the dependent growth curve.

[0149] In an implementation, the equation construction unit is specifically configured to: expand the pseudo-dynamic differential equation by using a truncated Legendre polynomial as a base function, wherein the truncated Legendre polynomial is used to describe a nonlinear interaction effect between the microorganisms; and decouple the expanded pseudo-dynamic differential equation by estimating variables and parameters of the expanded pseudo-dynamic differential equation, to obtain a plurality of independent regression equations, wherein each regression equation can be used to independently calculate an expression amount of a single microorganism.

[0150] In an implementation, a mathematical expression of the pseudo-dynamic differential equation after variable selection comprises:

[0151]

[0152] wherein i = 1, …, m; j = 1, …, n, m represents a total number of samples, and n represents a total number of microorganism OTUs, the natural growth curve of the OTUj, the allometric growth curve of the OTUj, the allometric growth curve of the OTUj in the ith sample, g ji the abundance level of the OTUj in the ith sample, E i the abundance level of all microbial OTU in the ith sample, j' is the OTU other than the OTUj, the dependent growth curve, representing the effect of the j'th OTU on the jth OTU, the allometric growth curve of the OTUj'(j'≠j) in the ith sample, the abundance level of the OTUj'(j'≠j) in the ith sample, d represents the number of OTUs having an effect on the OTUj, Θ j and Θ j←j' are equation parameters.

[0153] In an implementation, the network constructing unit is further configured to: calculate eigenvector centrality of each node in the interaction network; measure the influence of the significant variable corresponding to each node by using the eigenvector centrality; and determine the significant variable with the largest influence in the interaction network as the core microorganism.

[0154] In an implementation, the mathematical expression of the step of calculating the eigenvector centrality of each node in the interaction network comprises:

[0155]

[0156] wherein, x v is the eigenvector centrality of the node v, λ is the largest eigenvalue of the adjacency matrix of the interaction network, t∈M(v) is all nodes adjacent to the node v in the interaction network, x t is the eigenvector centrality of the node t, V is all nodes of the interaction network, a v,t is used to represent whether the nodes v and t are adjacent, if the node v is adjacent to the node t, a v,t = 1, if the node v is not adjacent to the node t, a v,t = 0.

[0157] The application further provides an interactive network construction device, which can include a processor and a memory. The processor and the memory are connected through a bus. The processor is configured to execute a plurality of instructions, and the memory is configured to store the plurality of instructions, which are suitable for being loaded and executed by the processor to implement the aforementioned interactive network construction method, in particular, to implement any of the implementable manners of the aforementioned interactive network construction method. The processor can be an electronic control unit (ECU), a central processing unit (CPU), a general-purpose processor, a coprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The processor can also be a combination of computing functions, such as one or more microprocessor combinations, a combination of a DSP and a microprocessor, and the like. In the embodiment, the processor can be a single-chip microcomputer, and various control functions can be realized by programming the single-chip microcomputer. The processor has the advantages of powerful computing capability and fast processing, and is configured to execute the functions of any of the initialization unit, the updating unit, and the output unit of the interactive network construction device in the aforementioned embodiments.

[0158] Specifically, the processor is configured to execute the functions of the equation construction unit, to construct a pseudo-dynamic differential equation and decouple the pseudo-dynamic differential equation into a plurality of independent regression equations, wherein the pseudo-dynamic differential equation includes a natural growth curve of microorganisms and a dependent growth curve of other microorganisms on the microorganisms; the functions of the equation decoupling unit, to solve the regression equations by using an adptive grouplasso algorithm to obtain independent expression amounts of each microorganism and at least one dependent expression amount between the microorganisms, and to screen out significant variables in the microorganisms; and the functions of the network construction unit, to construct a dynamic interactive network according to the independent expression amounts and the at least one dependent expression amount of the significant variables.

[0159] In an implementation, the processor is specifically configured to: determine the overall growth type of the significant variable in different samples, and the influence type and strength of other microorganisms on the significant variable according to the independent expression and dependent expression of the significant variable in different samples, wherein the overall growth type includes rising and falling, and the influence type includes inhibition and promotion; obtain the dynamic interaction network between the significant variables by using nodes and edges to represent the interaction relationship between the significant variables and other significant variables, and by assigning different icon styles to distinguish the significant variables of different overall growth types, and assigning different icon styles to distinguish the interaction relationship of different influence types and strengths.

[0160] In an implementation, the processor is specifically configured to: obtain the abundance level of each microorganism in different samples through the accumulation fermentation of a plurality of samples; establish a natural growth curve of the independent expression of the microorganism with the abundance level of the microorganism, and a dependent growth curve of the dependent expression of the microorganism with the abundance level of other microorganisms, to obtain a quasi-dynamic differential equation including the natural growth curve and the dependent growth curve; and expand the quasi-dynamic differential equation into a truncated series by using a Legendre polynomial as a base function, and perform parameter estimation on the coefficients of the Legendre polynomials in the expanded quasi-dynamic differential equation.

[0161] In an implementation, the processor is specifically configured to: expand the quasi-dynamic differential equation by using a truncated Legendre polynomial as a base function, wherein the truncated Legendre polynomial is used to characterize the nonlinear interaction effect between the microorganisms; and decouple a plurality of independent regression equations by estimating the variables and parameters of the expanded quasi-dynamic differential equation, wherein each regression equation can be used to independently calculate the expression of a single microorganism.

[0162] In an implementation, the mathematical expression of the quasi-dynamic differential equation after variable selection includes:

[0163]

[0164] wherein i = 1, …, m; j = 1, …, n, m represents the total number of samples, and n represents the total number of microbial OTUs, is the natural growth curve of the OTU, is the allometric growth curve of the OTUj, is the allometric growth curve of the OTUj in the ith sample, g ji is the abundance level of the OTUj in the ith sample, E i is the abundance level of all microbial OTUs in the ith sample, and j' is an OTU other than the OTUj, is the dependent growth curve, representing the influence of the j'th OTU on the jth OTU, is the growth rate of OTUj in the ith sample, is the abundance level of OTUj' (j'≠j) in the ith sample, d represents the number of OTUs that have an impact on OTUj, Θ j and Θ j←j' are equation parameters.

[0165] In an implementation, the processor is further configured to: calculate eigenvector centrality of each node in the interaction network; measure influence of the significant variable corresponding to each node by using the eigenvector centrality; and determine the significant variable with the largest influence in the interaction network as the core microorganism.

[0166] In an implementation, the mathematical expression of the step of calculating the eigenvector centrality of each node in the interaction network comprises:

[0167]

[0168] wherein x v is the eigenvector centrality of node v, λ is the largest eigenvalue of the adjacency matrix of the interaction network, t∈M(v) is all nodes adjacent to node v in the interaction network, x t is the eigenvector centrality of node t, V is all nodes in the interaction network, a v,t is used to represent whether nodes v and t are adjacent, a v,t = 1 if nodes v and t are adjacent, and a v,t = 0 if nodes v and t are not adjacent.

[0169] In an embodiment, the present application further provides a computer readable storage medium, which stores a plurality of instructions, the instructions being adapted to be loaded and executed by a processor to implement the method in any of the above embodiments. The processor is configured to execute the plurality of instructions, and the memory is configured to store the plurality of instructions, which are loaded and executed by the processor to implement the method of constructing the interaction network in the above embodiments.

[0170] Any combination of the technical features in the above embodiments can be made, and to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the description.

[0171] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be noted that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the patent protection scope of the present application should be subject to the appended claims.

Claims

1. A method for constructing an interaction network of microorganisms undergoing stacking fermentation, characterized in that, include: A quasi-dynamic differential equation is constructed and decoupled into multiple independent regression equations. The quasi-dynamic differential equation includes the natural growth curve of microorganisms and the growth curves dependent on the influence of other microorganisms on it. The adaptive group lasso algorithm was used to solve the regression equation to obtain the independent expression level of each microorganism and at least one dependent expression level with other microorganisms, and to screen out significant variables in the microorganisms. A dynamic interaction network is constructed based on the independent expression levels of significant variables and at least one dependent expression level.

2. The method according to claim 1, characterized in that, The step of constructing a dynamic interaction network based on the independent expression levels and at least one dependent expression level of a significant variable includes: The overall growth type of significant variables in different samples was determined based on their independent and dependent expression levels, as well as the types and strengths of their effects on other microorganisms. The overall growth type included increase and decrease, and the effect type included inhibition and promotion. Nodes and edges represent the interactions between significant variables and other significant variables, respectively. By assigning different icon styles to significant variables with different overall growth types, and different icon styles to distinguish interactions with different types and strengths of influence, a dynamic interaction network between significant variables is obtained.

3. The method according to claim 1, characterized in that, Constructing quasi-dynamic differential equations includes: Abundance levels of various microorganisms involved in stacked fermentation were obtained from multiple samples in different samples. By establishing the natural growth curve of the independent expression level of microorganisms as a function of the abundance level of microorganisms, and the dependent growth curve of the expression level of microorganisms as a function of the abundance level of other microorganisms, a quasi-dynamic differential equation including the natural growth curve and the dependent growth curve is obtained.

4. The method according to claim 1, characterized in that, Decoupling the quasi-dynamic differential equation into multiple independent regression equations includes: The quasi-dynamic differential equations are expanded using truncated Legendre polynomials as basis functions, wherein the truncated Legendre polynomials are used to characterize the nonlinear interaction effects between microorganisms. By estimating the variables and parameters of the expanded quasi-dynamic differential equation, multiple independent regression equations are obtained through decoupling. Each regression equation can be used to independently calculate the expression level of a single microorganism.

5. The method according to claim 1, characterized in that, Quasi-dynamic differential equations include: Where i = 1, ..., m; j = 1, ..., n, m represents the total number of samples, and n represents the total number of microbial OTUs. This is the natural growth curve of OUT. The allometric growth curve of OTUj is shown. Let g be the allometric growth curve of OTUj in sample i. ji E represents the abundance level of OTUj in the i-th sample. i Let j represent the abundance level of all microbial OTUs in sample i. ' For OTUs other than OTUj, The dependent growth curve represents the j-th... ' The impact of one OTU on the j-th OTU For the i-th sample, OTUj ' (j ' The allometric growth curve of ≠j) For the i-th sample, OTUj ' (j ' The abundance level of ≠j), where d represents the number of OTUs that influence OTUj, and Θ j and Θ j←j' These are the equation parameters.

6. The method according to claim 1, characterized in that, After constructing a dynamic interaction network based on the independent expression levels and at least one dependent expression level of significant variables, the method further includes: Calculate the feature vector centrality of each node in the interaction network; The influence of significant variables corresponding to each node is measured using the eigenvector centrality. The most influential significant variable in the interaction network was identified as the core microorganism.

7. The method according to claim 6, characterized in that, The mathematical expression for the step of calculating the eigenvector centrality of each node in the interaction network includes: Where, x v Let λ be the eigenvector centrality of node v, λ be the largest eigenvalue of the adjacency matrix of the interaction network, t∈M(v) be all nodes adjacent to node v in the interaction network, and x be the eigenvector centrality of node v. t Let be the eigenvector centrality of node t, V be all nodes in the interaction network, and a v,t This is used to indicate whether nodes v and t are adjacent. If node v is adjacent to node t, then a v,t =1, if node v and node t are not adjacent, then a v,t =0.

8. An interactive network construction device, characterized in that, The device includes: The equation building unit is used to construct quasi-dynamic differential equations and decouple them into multiple independent regression equations. The quasi-dynamic differential equations include the natural growth curve of microorganisms and the growth curves dependent on the influence of other microorganisms on them. The equation decoupling unit is used to solve the regression equation using the adaptive group lasso algorithm to obtain the independent expression level of each microorganism and at least one dependent expression level with other microorganisms, and to screen out significant variables in microorganisms. Network building blocks are used to construct dynamic interaction networks based on the independent expression levels of significant variables and at least one dependent expression level.

9. An interactive network construction device, characterized in that, The interactive network construction device includes a processor and a memory, which are connected via a bus; the processor is used to execute multiple instructions; the memory is used to store the multiple instructions, which are adapted to be loaded by the processor and executed as the interactive network construction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor and executing the interactive network construction method of any one of claims 1 to 7.

Images