Phenotype prediction and evaluation system for sugarcane filial generation

Abstract

The invention relates to the technical field of intelligent breeding, and discloses a sugarcane filial generation phenotype prediction and evaluation system, which comprises a genetic effect decomposition module, a metabolic pathway constraint module, a dynamic sensitivity bridging matrix construction module and a dynamic constraint optimization module which are sequentially executed and mutually coupled, wherein the metabolic pathway constraint module is used for executing flux equilibrium analysis based on a sugarcane core metabolic network and upper and lower bounds of enzyme activity mapped by parent QTL (Quantitative Trait Loci), generating an optimal flux solution of parents and constructing a filial generation flux feasible region; and solving an optimal phenotype trajectory by adopting sequential quadratic programming and calculating a comprehensive evaluation index. In the invention, a three-order epistatic tensor is constructed through a genetic effect decomposition module, tensor decomposition is carried out, an epistatic contribution vector of a parent is extracted, and a hybrid progeny flux feasible region is generated by combining flux equilibrium analysis of a metabolic pathway constraint module, so that a system can quantify a non-additive genetic effect and metabolic constraint before hybridization; and phenotype prediction is separated from field measured data dependence.

Description

Technical Field

[0001] This invention relates to the field of intelligent breeding technology, and in particular to a phenotypic prediction and evaluation system for sugarcane hybrid offspring. Background Technology

[0002] In the field of sugarcane hybrid breeding, genetic improvement of target traits such as high yield, high sugar content, and disease resistance is achieved through parental selection and offspring screening. The breeding process relies on quantitative genetic methods to evaluate parental combining ability and progeny typology.

[0003] In existing technologies, phenotypic prediction of sugarcane hybrid offspring mainly employs statistical models based on additive-dominant effects. This model first collects measured data on multiple traits, such as plant height, stalk diameter, and sugar content, from both parents under field conditions, and calculates broad-sense and narrow-sense heritability through analysis of variance. Subsequently, based on the Griffing diallel cross design or the North Carolina Design II experimental layout, a linear prediction equation is constructed, and the general combining ability and specific combining ability are estimated using the least squares method. Finally, selection indices are used to rank and screen the phenotypic offspring of the virtual hybrid combinations.

[0004] However, in the process of realizing the above technical solution, the inventors of this application found that the above technology has at least the following technical problems: the existing statistical model can only capture additive and dominant genetic effects, and cannot systematically quantify the interaction between non-additive effects such as epistasis and metabolic pathway constraints, which means that the prediction must rely on field measurement data after the offspring mature, and cannot achieve early phenotypic assessment without field trials before hybridization. Summary of the Invention

[0005] To overcome the above shortcomings, this invention provides a phenotypic prediction and evaluation system for sugarcane hybrid offspring, aiming to improve the problem that existing statistical models can only capture additive and dominant genetic effects and cannot achieve early phenotypic evaluation before hybridization without field trials.

[0006] In a first aspect, the present invention provides the following technical solution: a phenotypic prediction and evaluation system for sugarcane hybrid offspring, comprising a genetic effect decomposition module, a metabolic pathway constraint module, a dynamic sensitivity bridging matrix construction module, and a dynamic constraint optimization module, which are executed sequentially and coupled to each other, wherein:

[0007] The genetic effect decomposition module constructs a third-order epistatic tensor based on the QTL genotyping data of the parents and performs tensor decomposition to extract the epistatic contribution vector of the parents. At the same time, it calculates the general combining ability of the parents and the specific combining ability of the combination.

[0008] The metabolic pathway constraint module performs flux balance analysis based on the sugarcane core metabolic network and the upper and lower bounds of enzyme activity mapped by the parental QTLs, generates the optimal flux solution for both parents, and constructs the flux feasible region for the hybrid offspring.

[0009] The dynamic sensitivity bridging matrix construction module calculates instantaneous sensitivity and constructs a time-varying sensitivity bridging matrix by applying small perturbations to key fluxes and recording phenotypic parameter responses.

[0010] The dynamic constraint optimization module models the phenotypic profile as a family of piecewise logistic curves. The upper-level perturbation is passed to the phenotypic parameters through a time-varying sensitivity bridging matrix. Combined with the metabolic weight function and flux feasible region constraints, the optimal phenotypic trajectory is solved by sequential quadratic programming, and the comprehensive evaluation index is calculated.

[0011] Preferably, the genetic effect decomposition module constructs a third-order epistatic tensor and performs tensor decomposition to extract the epistatic contribution vectors of the parents, including:

[0012] Construct a third-order epitope tensor based on the QTL typing data of the parent model;

[0013] Perform CANDECOMP / PARAFAC decomposition on the third-order hypertropic tensor to separate the factor vectors;

[0014] Extract the epistatic contribution vector of the parents from the factor vectors obtained from the decomposition.

[0015] Preferably, the process by which the metabolic pathway constraint module generates the optimal flux solution for both parents and constructs the flux feasible region for the hybrid offspring includes:

[0016] Construct a core metabolic network for sugarcane that includes multiple biochemical reactions;

[0017] Map parental QTLs to upper and lower bounds of metabolic response flux;

[0018] Perform flux balance analysis on each parent to obtain the optimal flux solution;

[0019] Convex hulls of hybrid offspring are constructed based on the optimal flux solutions of the two parents to create a flux feasible region.

[0020] Preferably, the process by which the dynamic sensitivity bridging matrix construction module calculates instantaneous sensitivity and constructs a time-varying sensitivity bridging matrix includes:

[0021] Apply small perturbations to each key flux;

[0022] Record the temporal changes of phenotypic parameters after perturbation;

[0023] Calculate the instantaneous sensitivity of phenotypic parameters to flux disturbances;

[0024] A time-varying sensitivity bridging matrix is ​​constructed based on instantaneous sensitivity and then smoothed.

[0025] Preferably, the process by which the dynamic constraint optimization module transfers the hyperordinate perturbation to the phenotypic parameters via the time-varying sensitivity bridging matrix includes:

[0026] The epistatic contribution vector of the parents is mapped to metabolic flux perturbation;

[0027] Metabolic flux perturbations are transferred to phenotypic parameters through a time-varying sensitive bridging matrix;

[0028] The transmitted perturbation is superimposed on the baseline phenotypic parameters based on general and special cooperating forces.

[0029] Preferably, the process by which the dynamic constraint optimization module combines the metabolic weight function and the flux feasible region constraint includes:

[0030] The metabolic weighting function is constructed based on the normalized position of the current flux relative to the upper and lower bounds.

[0031] Apply the metabolic weighting function to the regulation term of the piecewise logistic curve family;

[0032] In the optimization, flux feasible region constraints and lower bound constraints on metabolic weight function are applied.

[0033] Preferably, the process by which the dynamic constraint optimization module solves for the optimal phenotypic trajectory and calculates the comprehensive evaluation index using sequential quadratic programming includes:

[0034] Construct an optimization objective that includes the deviation between the phenotypic trajectory and the target curve, as well as penalties for hierarchical perturbations;

[0035] Perform sequential quadratic programming under the constraints of flux feasible region, metabolic weight function, and epistatic perturbation propagation;

[0036] Generate a full reproductive period phenotypic trajectory based on the solved phenotypic parameters;

[0037] The phenotypic trajectory is weighted and integrated to calculate the comprehensive evaluation index.

[0038] Preferably, the system inputs are the QTL genotyping data of the two parents, the sugarcane core metabolic network topology, and the breeding target curve.

[0039] Preferably, the system output includes the full-life-cycle phenotypic trajectory of the hybrid offspring and a comprehensive evaluation index.

[0040] Ideally, the four modules are coupled to each other through a closed-loop link: from the epistatic contribution vector to metabolic flux perturbation, from the time-varying sensitivity bridging matrix to phenotypic parameter transfer, and from the metabolic weight function to phenotypic trajectory regulation.

[0041] The present invention has the following beneficial effects:

[0042] 1. In this invention, a third-order epistatic tensor is constructed and decomposed by a genetic effect decomposition module to extract the epistatic contribution vectors of the two parents. Combined with the flux balance analysis of the metabolic pathway constraint module, the flux feasible region of the hybrid offspring is generated, so that the system can quantify non-additive genetic effects and metabolic constraints before hybridization, and realize phenotypic prediction without dependence on field measured data.

[0043] 2. In this invention, the dynamic sensitivity bridging matrix construction module applies small perturbations to key fluxes and records the phenotypic parameter responses, calculates instantaneous sensitivity, and constructs a time-varying sensitivity bridging matrix. This transmits metabolic perturbations to phenotypic parameters in real time. Combined with the sequential quadratic programming of the dynamic constraint optimization module, it generates a phenotypic trajectory throughout the entire growth period, thus completing the continuous prediction from static combining ability to dynamic growth process.

[0044] 3. In this invention, the four modules are coupled through a closed-loop link from the epistatic contribution vector to metabolic flux perturbation, from the time-varying sensitivity bridging matrix to phenotypic parameter transfer, and from the metabolic weight function to phenotypic trajectory adjustment. This supports batch parallel computation of multiple hybridization combinations and outputs a comprehensive evaluation index, which can be directly used for virtual screening and ranking. Attached Figure Description

[0045] Figure 1 This is a diagram of the phenotypic prediction and evaluation system for sugarcane hybrid offspring proposed in this invention.

[0046] Figure 2 This is a flowchart of the genetic effect decomposition module of a phenotypic prediction and evaluation system for sugarcane hybrid offspring proposed in this invention.

[0047] Figure 3 This is a flowchart of the metabolic pathway constraint module of a phenotypic prediction and evaluation system for sugarcane hybrid offspring proposed in this invention.

[0048] Figure 4 This is a flowchart of the dynamic constraint optimization module of a sugarcane hybrid progeny phenotypic prediction and evaluation system proposed in this invention. Detailed Implementation

[0049] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Example 1

[0051] Reference Figures 1-4In a first embodiment of the present invention, the present invention provides a phenotypic prediction and evaluation system for sugarcane hybrid offspring, comprising a genetic effect decomposition module, a metabolic pathway constraint module, a dynamic sensitivity bridging matrix construction module, and a dynamic constraint optimization module, which are executed sequentially and coupled to each other, wherein:

[0052] The genetic effect decomposition module constructs a third-order epistatic tensor based on the QTL genotyping data of the parents and performs tensor decomposition to extract the epistatic contribution vector of the parents. At the same time, it calculates the general combining ability of the parents and the specific combining ability of the combination.

[0053] The metabolic pathway constraint module performs flux balance analysis based on the sugarcane core metabolic network and the upper and lower bounds of enzyme activity mapped by the parental QTLs, generates the optimal flux solution for both parents, and constructs the flux feasible region for the hybrid offspring.

[0054] The dynamic sensitivity bridging matrix construction module calculates instantaneous sensitivity and constructs a time-varying sensitivity bridging matrix by applying small perturbations to key fluxes and recording phenotypic parameter responses.

[0055] The dynamic constraint optimization module models the phenotypic profile as a family of piecewise logistic curves. The upper-level perturbation is passed to the phenotypic parameters through a time-varying sensitivity bridging matrix. Combined with the metabolic weight function and flux feasible region constraints, the optimal phenotypic trajectory is solved by sequential quadratic programming, and the comprehensive evaluation index is calculated.

[0056] Specifically, the system comprises a genetic effect decomposition module, a metabolic pathway constraint module, a dynamic sensitivity bridging matrix construction module, and a dynamic constraint optimization module. These four modules are connected sequentially via a data interface. Data is transmitted sequentially from the genetic effect decomposition module to the metabolic pathway constraint module, and then via the dynamic sensitivity bridging matrix construction module to the dynamic constraint optimization module. The output of the dynamic constraint optimization module is fed back to the metabolic pathway constraint module, forming a closed-loop data flow. System inputs include QTL genotyping data of the two parents, the stoichiometric matrix of the sugarcane core metabolic network, and the breeding target curve. System outputs include the plant height trajectory, stalk diameter trajectory, sugar content trajectory, and comprehensive evaluation index of the hybrid offspring throughout the entire growth period.

[0057] The genetic effects decomposition module receives QTL genotyping data from both parents and calculates the epistatic contribution vector, general combining ability, and specific combining ability. The metabolic pathway constraint module receives the QTL data and stoichiometry matrix output from the genetic effects decomposition module, performs flux balance analysis, and outputs the flux feasible region for the hybrid offspring. The dynamic sensitivity bridging matrix construction module receives the key fluxes output from the metabolic pathway constraint module and calculates the time-varying sensitivity bridging matrix. The dynamic constraint optimization module receives the time-varying sensitivity bridging matrix, epistatic contribution vector, flux feasible region, and breeding target curve, performs sequence quadratic programming, and outputs the phenotypic trajectory and comprehensive evaluation index.

[0058] The system adopts a modular program structure, with each module transmitting data through defined structure variables. The genetic effect decomposition module and the metabolic pathway constraint module are connected via a structure containing epistatic contribution vectors. The metabolic pathway constraint module and the dynamic sensitivity bridging matrix construction module are connected via an array containing key flux values. The dynamic sensitivity bridging matrix construction module and the dynamic constraint optimization module are connected via a two-dimensional array containing time-varying sensitivity bridging matrices.

[0059] During system execution, the genetic effect decomposition module is executed first to generate epistatic contribution vectors and combining ability parameters. Then, the metabolic pathway constraint module is activated to generate the flux feasible region. Next, the dynamic sensitivity bridging matrix construction module is run to generate the time-varying sensitivity bridging matrix. Finally, the dynamic constraint optimization module is activated to solve for the phenotypic trajectory. During iteration, the dynamic constraint optimization module calls the metabolic pathway constraint module to verify the flux feasible region constraints. The system supports batch processing of multiple hybridization combinations. Input data is stored in batches in a two-dimensional array, with each combination corresponding to one row. Output results are stored in the same order in the results file.

[0060] Furthermore, the genetic effect decomposition module constructs a third-order epistatic tensor and performs tensor decomposition to extract the epistatic contribution vectors of the parents, including:

[0061] Construct a third-order epitope tensor based on the QTL typing data of the parent model;

[0062] Perform CANDECOMP / PARAFAC decomposition on the third-order hypertropic tensor to separate the factor vectors;

[0063] Extract the epistatic contribution vector of the parents from the factor vectors obtained from the decomposition.

[0064] Specifically, the genetic effect decomposition module receives QTL genotyping data from both parents as input. The QTL genotyping data is stored in matrix form, with rows corresponding to QTL loci and columns corresponding to the two parents. Matrix elements take values ​​of 0 or 2, representing homozygous reference alleles, heterozygous alleles, or homozygous alternative alleles, respectively. The total number of loci is denoted as...

[0065] The module first performs standardization on the QTL genotyping data. Standardization involves subtracting the mean of that locus in the parental population from the genotyping value of each locus, and then dividing by the standard deviation to obtain the standardized genotyping matrix Z, with dimensions [missing information]. The construction of the third-order hypertropic tensor T is based on the normalized fractal matrix Z. The dimension of T is... Initialize to a zero tensor. For tensor elements... The calculation formula is:

[0066] ;

[0067] in, Traverse 1 to Indicates the first parent in the first year Standardized values ​​of the locus, Indicates the second parent in the first The standardized value of the site. This calculation is performed through a triple loop, with the loop index executed sequentially.

[0068] Tensor decomposition uses the CANDECOMP / PARAFAC method. Decomposition rank... Set to predefined values, ranging from 3 to 10. The decomposition process uses alternating least squares. Initialize the factor matrix. All dimensions are The element randomly takes a value in the interval [-1, 1].

[0069] The iterative steps are as follows:

[0070] S1, Fixed And C, update A.

[0071] The r-th column is minimized Obtain, among which For tensor The pattern-1 expansion matrix, This represents the Khatri-Rao product. Update using the least squares solution:

[0072] ;

[0073] S2: Fix A and C, update B, the calculation method is similar to S1.

[0074] S3: Fix A and B and update C. The calculation method is similar to S1.

[0075] S4. Calculate reconstruction error

[0076] S5, when Stop if the value is less than the threshold of 0.001 or the number of iterations reaches 1000; otherwise, return to S1.

[0077] After decomposition, the factor matrix The first row corresponds to the epistatic contribution vector of the first parent. , dimension is Factor matrix The first row corresponds to the epistatic contribution vector of the second parent. , dimension is The calculation of general combining ability and specific combining ability is based on parental historical phenotypic data. Historical phenotypic data includes plant height, stalk diameter, and sugar content, stored in a vector... and In the middle, the length is the number of traits. .

[0078] General cooperation force The calculation is as follows: ;in, This represents the overall mean of all parental phenotypes. Specific combining ability. The calculation is as follows: ,in, This represents the phenotypic mean of historical hybridization combinations. If no historical hybridization data is available, then... Set it to 0.

[0079] The module output includes a hyperordinate contribution vector. General cooperating force Special cooperation ability The output data is encapsulated in a structure, which is passed to the metabolic pathway constraint module via a pointer.

[0080] Furthermore, the process by which the metabolic pathway constraint module generates the optimal flux solution for both parents and constructs the flux feasible region for the hybrid offspring includes:

[0081] Construct a core metabolic network for sugarcane that includes multiple biochemical reactions;

[0082] Map parental QTLs to upper and lower bounds of metabolic response flux;

[0083] Perform flux balance analysis on each parent to obtain the optimal flux solution;

[0084] Convex hulls of hybrid offspring are constructed based on the optimal flux solutions of the two parents to create a flux feasible region.

[0085] Specifically, the metabolic pathway constraint module receives QTL genotyping data output from the genetic effect decomposition module, as well as a predefined stoichiometric matrix of the sugarcane core metabolic network. (Stoichiometric matrix) Dimensions ,in, For the amount of metabolites, Number of biochemical reactions. Matrix elements. Indicates the first The metabolite in the first In a reaction, the stoichiometric coefficients are positive for products, negative for substrates, and zero for no participation.

[0086] The module first loads the sugarcane core metabolic network. This network includes pathways such as photosynthetic carbon fixation, sucrose synthesis, and cellulose deposition, totaling 20 reactions and 95 metabolites. After importing, it is converted into a stoichiometric matrix. and reaction reversibility vector ,in, Indicates reaction reversible, It indicates that it is irreversible.

[0087] The mapping from QTLs to flux upper and lower bounds is based on a site-enzyme mapping table. This table is a two-dimensional array, with rows corresponding to QTL sites and columns containing the enzyme code, influence direction, and influence strength. The influence direction takes values ​​of 1 (upregulation) or -1 (downregulation). For each parent, its QT1 genotypes are iterated through:

[0088] S1. When the site type is 2 (homozygous substitution), the flux limit of the corresponding enzyme is adjusted according to the intensity.

[0089] S2, upper limit of flux for upregulated enzymes Increase according to the following formula:

[0090] ;

[0091] The Lower World Keep

[0092] S3, lower limit of enzyme flux reduction Reduce using the following formula:

[0093] ;

[0094] Upper Realm Keep .

[0095] S4, Initial Boundary Read from the network standard value, the lower bound of irreversible reaction is set to 0.

[0096] The adjusted bounds are stored in the matrix. (first parent) and (Second parent), all dimensions are... Flux balance analysis was performed independently for each parent. The optimization objective was the biomass synthesis reaction flux, with a target coefficient vector. The biomass response is set to 1, and all others to 0. The constraints are:

[0097] ;

[0098] ;

[0099] The solution is obtained using linear programming functions from the COBRA toolbox. Irreversible reaction constraint. The optimal flux vector of the first parent is obtained by solving the problem. The second parent's optimal flux vector All dimensions are Flux Feasibility Region of Hybrid Progeny Convex hull construction is performed using the optimal flux vectors of the two parents. The steps for calculating the convex hull are as follows:

[0100] S1, Construction Matrix Dimension .

[0101] S2, for each reaction Take the lower bound:

[0102] ;

[0103] Upper Realm:

[0104] ;

[0105] S3, Generation Boundary Vector This forms a rectangular feasible region approximation.

[0106] Key flux screening from Extracting phenotypic-related data from each reaction The flux was selected based on the absolute value of the Pearson correlation coefficient between the flux and historical phenotypic data. The coefficient of variation of flux for each reaction was calculated, and the top-ranked fluxes were selected using a correlation coefficient threshold of 0.3. The reaction. Key flux value from Extract from and form a vector The critical flux feasible region for hybrid offspring is... Elements are compared one by one, and the module output includes the flux feasible region. Key flux benchmark The output data is encapsulated in a structure. The structure is passed to the dynamic sensitivity bridging matrix construction module via a pointer.

[0107] Furthermore, the process by which the dynamic sensitivity bridging matrix construction module calculates instantaneous sensitivity and constructs a time-varying sensitivity bridging matrix includes:

[0108] Apply small perturbations to each key flux;

[0109] Record the temporal changes of phenotypic parameters after perturbation;

[0110] Calculate the instantaneous sensitivity of phenotypic parameters to flux disturbances;

[0111] A time-varying sensitivity bridging matrix is ​​constructed based on instantaneous sensitivity and then smoothed.

[0112] Specifically, the dynamic sensitivity bridging matrix construction module receives key flux benchmarks output by the metabolic pathway constraint module. , dimension is And a predefined initial set of phenotypic parameters. The initial set of phenotypic parameters includes logistic curve parameters for plant height, stalk diameter, and sugar content, respectively. ,total Each parameter is stored in a vector. The module performs perturbation experiments using a discrete-time grid. The time grid is defined as follows: ,in, . Covering the entire reproductive period of 360 days. Disturbance amplitude. Set as the critical flux baseline ,Right now .

[0113] The disturbance process Execute each key throughput one by one:

[0114] S1, Loading reference flux .

[0115] S2, for the first A key flux is used to construct a perturbation vector. ,in, For the first A standard unit vector of dimension 1.

[0116] S3, will The reverse mapping is performed to the complete flux space. The mapping uses linear interpolation: perturbed flux. If the reaction Critical flux Control, or keep .

[0117] S4, Execute steady state Verification in the feasible region If the boundary is exceeded, project along the boundary to the nearest feasible point.

[0118] The time-series response of phenotypic parameters after perturbation was calculated using a metabolism-phenotypic function. This function is based on the flux-to-growth rate conversion: ,in, These are empirical conversion coefficients. Parameters after perturbation. Obtained through numerical integration of the logistic equation:

[0119] ;

[0120] The remaining parameters are similar to integrals. Unperturbed reference response. Calculate in the same way.

[0121] Instantaneous sensitivity at each time point calculate:

[0122] ;

[0123] in, Calculate the traversal of all The results are stored in a three-dimensional array. In the middle, dimension The construction of the time-varying sensitivity bridging matrix starts from... Extract each time slice:

[0124] S1, For fixed Extracting a matrix Dimension .

[0125] S2, to Perform outlier removal: when Then set it as the median of that column.

[0126] S3. Store all time slices to form a time-varying matrix set.

[0127] Smoothing is achieved using cubic spline interpolation. The interpolation is performed independently of each time dimension. implement:

[0128] S1, Input discrete points

[0129] S2, Boundary conditions are set as natural splines, and the second derivative is... and The value is zero.

[0130] S3, Interpolation generates continuous functions The sampling interval is 1 day, and the smoothing matrix is ​​obtained.

[0131] S4. Store as a three-dimensional tensor, dimension... .

[0132] The module output is a smoothed time-varying sensitivity bridging matrix. The data is stored in HDF5 format and includes the time axis and matrix values. The file pointer is passed to the dynamic constraint optimization module.

[0133] Furthermore, the process by which the dynamic constraint optimization module transmits the higher-order perturbation to the phenotypic parameters via the time-varying sensitivity bridging matrix includes:

[0134] The epistatic contribution vector of the parents is mapped to metabolic flux perturbation;

[0135] Metabolic flux perturbations are transferred to phenotypic parameters through a time-varying sensitive bridging matrix;

[0136] The transmitted perturbation is superimposed on the baseline phenotypic parameters based on general and special cooperating forces.

[0137] The dynamic constraint optimization module combines the metabolic weight function and flux feasible region constraints in the following process:

[0138] The metabolic weighting function is constructed based on the normalized position of the current flux relative to the upper and lower bounds.

[0139] Apply the metabolic weighting function to the regulation term of the piecewise logistic curve family;

[0140] In the optimization, flux feasible region constraints and lower bound constraints on metabolic weight function are applied.

[0141] The dynamic constraint optimization module uses sequential quadratic programming to solve for the optimal phenotypic trajectory and calculate the comprehensive evaluation index. The process includes:

[0142] Construct an optimization objective that includes the deviation between the phenotypic trajectory and the target curve, as well as penalties for hierarchical perturbations;

[0143] Perform sequential quadratic programming under the constraints of flux feasible region, metabolic weight function, and epistatic perturbation propagation;

[0144] Generate a full reproductive period phenotypic trajectory based on the solved phenotypic parameters;

[0145] The phenotypic trajectory is weighted and integrated to calculate the comprehensive evaluation index.

[0146] Specifically, the dynamic constraint optimization module receives the hypergenetic contribution vector output by the genetic effect decomposition module. general cooperation Special cooperation ability ;Receive the flux feasible region bound output by the metabolic pathway constraint module and key flux benchmarks ;Receive the time-varying sensitivity bridging matrix output by the dynamic sensitivity bridging matrix construction module The module simultaneously loads the breeding target curve. The time grid is

[0147] The mapping from epistatic perturbations to metabolic fluxes is achieved through a predefined mapping matrix. accomplish. Dimensions The QTL-enzyme correspondence table is expanded to the Kronecker product form. The perturbation vector is calculated as follows: ;in, For Kronecker product, generate Vector. The perturbation is constant throughout the entire time span.

[0148] Disturbance propagation uses a time-varying sensitive bridging matrix: ;in, Dimensions Baseline phenotypic parameters Obtained by linear interpolation of general and special mating forces: ;in, For the standard growth parameter vector, This is the dominant offset vector. The final phenotypic parameters are:

[0149] The metabolic weighting function is constructed based on the current position of the critical flux. Current flux pass Reverse mapping to Steady-state solution. Normalized position calculation is as follows: The weighting function is: ; The lower bound constraint is .

[0150] Phenotypic trajectory modeling is as follows:

[0151] ;

[0152] Optimization variables are At discrete time points The value on, The objective function is:

[0153] ;

[0154] in, The constraint set is:

[0155] ;

[0156] ;

[0157] ;

[0158] Sequential Quadratic Programming (SQP) uses the Snopt solver. The initial point is... Iteration steps:

[0159] S1. Linearize the constraints and quadratic approximation objectives at the current point.

[0160] S2. Solve the quadratic programming subproblem to obtain the search direction.

[0161] S3, perform a line search to determine the step size.

[0162] S4, Update

[0163] S5. When the gradient norm is less than Stop after 200 iterations.

[0164] After solving, the phenotypic trajectory is obtained through Interpolation generates a continuous curve. The comprehensive evaluation index is calculated as follows:

[0165] ;

[0166] in, The integral is obtained using the trapezoidal rule.

[0167] The module output includes the optimal phenotypic trajectory. and evaluation index The data is stored in NetCDF format. The file pointer is passed to the system output interface.

[0168] Furthermore, the system inputs are the QTL genotyping data of the two parents, the topology of the sugarcane core metabolic network, and the breeding target curve; the system outputs include the phenotypic trajectory of the hybrid offspring throughout their entire growth period and the comprehensive evaluation index; the four modules are coupled to each other through a closed-loop link from the epistatic contribution vector to metabolic flux perturbation, from the time-varying sensitivity bridging matrix to the phenotypic parameter transfer, and from the metabolic weight function to the phenotypic trajectory regulation.

[0169] Specifically, the system integration is implemented using C++, with all modules encapsulated as independent classes, and data interaction achieved through pointers and structures. The main program first instantiates four objects: the genetic effect decomposition module, the metabolic pathway constraint module, the dynamic sensitivity bridging matrix construction module, and the dynamic constraint optimization module.

[0170] The data flow interface uses a unified structure definition. The structure passed from the genetic effect decomposition module to the metabolic pathway constraint module includes the epistatic contribution vector field, the GCA field, and the SCA field. The structure passed from the metabolic pathway constraint module to the dynamic sensitivity bridging matrix construction module includes the flux lower bound array, the upper bound array, and the key flux baseline array. The structure passed from the dynamic sensitivity bridging matrix construction module to the dynamic constraint optimization module includes the time-varying sensitivity matrix three-dimensional array and the time axis array.

[0171] The closed-loop process is implemented using an iterative controller. After the dynamic constraint optimization module starts, the main program sets the maximum number of iterations to 5. At the end of each optimization iteration, the dynamic constraint optimization module sends the current flux vector back to the metabolic pathway constraint module. Upon receiving the returned flux, the metabolic pathway constraint module performs flux balance analysis to verify the feasible region constraints. If the flux exceeds the limit, it projects along the boundary and updates the key flux baseline, then re-triggers the dynamic sensitivity bridging matrix construction module to refresh the time-varying sensitivity bridging matrix. The updated bridging matrix is ​​then input back into the dynamic constraint optimization module to form a closed loop.

[0172] The parallel computing architecture is implemented based on OpenMP. The system supports processing multiple hybridization combinations simultaneously. Input QTL data is stored in batches in a two-dimensional container, with each row corresponding to one combination. The main program creates a thread pool, with the number of threads equal to the number of available CPU cores. Each thread independently executes the complete four-module process, and shared metabolic network topology data is protected between threads using mutex locks. Output results are merged into a single file in the order of input.

[0173] Fault tolerance is handled by setting data validation at the entry point of each module. The genetic effect decomposition module checks whether the QTL genotyping matrix dimension is L×2. The metabolic pathway constraint module checks whether the number of rows in the stoichiometry matrix matches the number of metabolites. The dynamic sensitivity bridging matrix construction module checks whether the flux remains within the feasible region after perturbation. If validation fails, the module records the error code, skips the current combination, and outputs the default zero trajectory.

[0174] When the system starts, the main program reads the configuration file to specify the input file path, output directory, and number of parallel threads. When the program ends, it releases all dynamically allocated memory, closes file handles, and outputs a log recording the total number of processed combinations and the number of errors.

[0175] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A system for phenotypic prediction and evaluation of sugarcane progeny from crosses, characterized by, The method comprises genetic effect decomposition module, metabolic pathway constraint module, dynamic sensitivity bridge matrix construction module and dynamic constraint optimization module which are executed sequentially and coupled with each other, wherein: The genetic effect decomposition module constructs a third-order epistatic tensor based on the QTL typing data of the parents and performs tensor decomposition to extract the epistatic contribution vectors of the parents, and calculates the general combining ability of the parents and the special combining ability of the combination; The metabolic pathway constraint module performs flux balance analysis based on the core metabolic network of sugarcane and the upper and lower bounds of enzyme activity mapped by the QTL of the parents to generate the optimal flux solution of the parents and construct the flux feasible region of the hybrid offspring; The dynamic sensitivity bridge matrix construction module calculates the instantaneous sensitivity by applying a small disturbance to the key flux and recording the response of the phenotypic parameter, and constructs the time-varying sensitivity bridge matrix; The dynamic constraint optimization module models the phenotype of the offspring as a family of piecewise logistic curves, transmits the epistatic disturbance to the phenotypic parameter through the time-varying sensitivity bridge matrix, combines the metabolic weight function and the flux feasible region constraint, and solves the optimal phenotype trajectory and calculates the comprehensive evaluation index by using sequential quadratic programming.

2. A system for phenotyping and evaluation of sugarcane progeny of crosses as claimed in claim 1, wherein, The process of the genetic effect decomposition module constructing a third-order epistatic tensor and performing tensor decomposition to extract the epistatic contribution vectors of the parents comprises: constructing a third-order epistatic tensor based on the QTL typing data of the parents; performing CANDECOMP / PARAFAC decomposition on the third-order epistatic tensor to separate the factor vectors; extracting the epistatic contribution vectors of the parents from the factor vectors obtained by decomposition.

3. A system for phenotypic prediction and evaluation of sugarcane progeny of a cross according to claim 1 or 2, wherein, The process of the metabolic pathway constraint module generating the optimal flux solution of the parents and constructing the flux feasible region of the hybrid offspring comprises: constructing a core metabolic network of sugarcane comprising a plurality of biochemical reactions; mapping the QTL of the parents to the flux upper and lower bounds of the metabolic reactions; performing flux balance analysis on each parent to obtain the optimal flux solution; constructing the flux feasible region of the hybrid offspring based on the convex hull of the optimal flux solutions of the parents.

4. A system for phenotyping and evaluation of sugarcane progeny of a cross according to claim 3, wherein, The process of the dynamic sensitivity bridge matrix construction module calculating the instantaneous sensitivity and constructing the time-varying sensitivity bridge matrix comprises: applying a small disturbance to the key flux one by one; recording the time sequence change of the phenotypic parameter after the disturbance; calculating the instantaneous sensitivity of the phenotypic parameter to the flux disturbance; constructing the time-varying sensitivity bridge matrix based on the instantaneous sensitivity and performing smoothing processing.

5. A system for phenotyping and evaluation of sugarcane progeny of a cross as claimed in claim 4, wherein, The process of the dynamic constraint optimization module transmitting the epistatic disturbance to the phenotypic parameter through the time-varying sensitivity bridge matrix comprises: mapping the epistatic contribution vectors of the parents to metabolic flux disturbance; transmitting the metabolic flux disturbance to the phenotypic parameter through the time-varying sensitivity bridge matrix; superimposing the transmitted disturbance to the baseline phenotypic parameter based on the general combining ability and the special combining ability.

6. A system for phenotypic prediction and evaluation of sugarcane progeny of a cross according to claim 5, wherein, The process of the dynamic constraint optimization module combining the metabolic weight function and the flux feasible region constraint comprises: constructing the metabolic weight function based on the normalized position of the current flux relative to the upper and lower bounds; applying the metabolic weight function to the adjustment term of the family of piecewise logistic curves; applying the flux feasible region constraint and the lower bound constraint of the metabolic weight function in the optimization.

7. A system for phenotypic prediction and evaluation of sugarcane progeny of a cross according to claim 6, wherein, The process of the dynamic constraint optimization module solving the optimal phenotype trajectory and calculating the comprehensive evaluation index by using sequential quadratic programming comprises: An optimization objective is constructed including the deviation of the phenotype trajectory from the target curve and the penalty of epistatic perturbation; Sequential quadratic programming is performed under the constraints of flux feasible region, metabolic weight function and epistatic perturbation transmission; The whole growth period phenotype trajectory is generated based on the solved phenotype parameters; The comprehensive evaluation index is calculated by weighted integration of the phenotype trajectory.

8. A system for phenotyping and evaluation of sugarcane progeny of a cross according to claim 7, wherein, The system inputs are QTL typing data of the parents, the topology of the core metabolic network of sugarcane and the breeding target curve.

9. A system for phenotyping and evaluation of sugarcane progeny of crosses as claimed in claim 1, wherein, The system outputs include the whole growth period phenotype trajectory and the comprehensive evaluation index of the hybrid offspring.

10. A system for phenotypic prediction and evaluation of sugarcane progeny of a cross according to any one of claims 1, wherein, The four modules are coupled through the closed loop link of epistatic contribution vector to metabolic flux perturbation, time-varying sensitivity bridge matrix to phenotype parameter transmission, metabolic weight function to phenotype trajectory adjustment.

Images