Breeding auxiliary method and system for red sea bream based on multi-trait coordinated selection

Abstract

The application discloses a kind of based on multi-trait collaborative selection redfin snapper breeding auxiliary method and system, comprising: phenotype data acquisition and gene detection are carried out, and standardization phenotype matrix and genotype matrix are constructed;Genetic parameter estimation and breeding value prediction are carried out based on standardization phenotype matrix and genotype matrix, and breeding value prediction information is obtained;The correlation between the traits of redfin snapper to be bred is analyzed and a trait correlation network is constructed, multi-trait dynamic weight configuration is carried out, and a dynamic weight set is obtained;According to breeding value prediction information and dynamic weight set, calculate genomic selection index and analyze parent priority, generate parent selection priority list;A candidate parent pool is generated through the parent selection priority list, mating combination analysis is carried out, and the genetic gain of the generated mating combination is predicted, and a mating assistance scheme is generated. Improve the genetic gain accuracy and balance of redfin snapper breeding, break through the limitations of traditional experience and improve breeding efficiency.

Description

Technical Field

[0001] This invention relates to the field of red snapper breeding technology, and in particular to a red snapper breeding auxiliary method and system based on multi-trait synergistic selection. Background Technology

[0002] As an important marine economic fish, the genetic improvement of redfin snapper has long relied on traditional breeding methods. Conventional breeding methods focus on single-trait directional selection, such as selecting individuals with fast growth rates for breeding. However, in actual production, it has been found that continuous intensification of single-trait selection can lead to the decline of other important traits. For example, populations with excessively high growth performance may experience decreased disease resistance and reduced feed conversion efficiency. This trait-related genetic antagonism is particularly pronounced in redfin snapper, where growth rate and stress resistance exhibit a negative genetic correlation, making it difficult for single-trait optimization strategies to achieve comprehensive genetic progress.

[0003] Traditional breeding strategies have long relied on single-trait directional selection, leading to pronounced genetic antagonism among key traits. This single-trait priority model not only limits the rate of overall genetic progress but also exacerbates the decline in population genetic diversity by neglecting trait synergy, while also posing a risk of inbreeding depression. Therefore, this paper proposes a breeding aid method and system for redfin snapper based on multi-trait synergistic selection to improve the environmental adaptability and breeding efficiency of redfin snapper breeding. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a breeding assistance method and system for redfin snapper based on multi-trait synergistic selection.

[0005] To achieve the above objectives, the first aspect of this invention provides a breeding assistance method for redfin snapper based on multi-trait synergistic selection, comprising:

[0006] Phenotypic data and gene testing were conducted on the redfin snapper population to be bred, and a standardized phenotypic matrix and genotype matrix were constructed based on the collected data.

[0007] Genetic parameters are estimated and breeding values ​​are predicted based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information.

[0008] Based on the breeding value prediction information, the correlation between traits of the redfin snapper to be bred is analyzed and a trait association network is constructed. Then, the dynamic weights of multiple traits are configured to obtain a dynamic weight set.

[0009] Based on the breeding value prediction information and dynamic weight set, the genomic selection index is calculated and the parental priority is analyzed to generate a parental selection priority list.

[0010] A candidate parent pool is generated by the parent selection priority list, and mating combination analysis is performed on the candidate parent pool. Genetic gain prediction is then performed on the generated mating combinations to generate mating assistance schemes.

[0011] In this scheme, the phenotypic data collection and gene testing of the redfin snapper population to be bred, and the construction of a standardized phenotypic matrix and genotype matrix based on the collected data, specifically include:

[0012] The weighing platform was used to weigh the red snapper population to be bred in the target breeding pond within a preset collection period. The body length growth rate of the red snapper population to be bred within the preset collection period was calculated by deploying a multi-source sensor array and combining machine vision technology.

[0013] The amount of uneaten feed in the target breeding pond is obtained by deploying uneaten feed collection devices. The difference between the uneaten feed and the daily feed amount is calculated to obtain the feed intake data. The feed intake-weight gain ratio is calculated using the obtained fish body weight data.

[0014] Based on a pre-set disease resistance breeding test plan, a disease resistance test was conducted on redfin snapper in the target breeding pond. The survival rate within the pre-set time period was recorded as the quantitative value of the disease resistance trait. Finally, the original phenotypic dataset was obtained by combining the fish body identification.

[0015] Subsequently, gene samples were collected from the redfin snapper population to be bred. Based on the collected gene samples, whole-genome typing was performed using gene detection technology to form an original genotype dataset containing locus genotypes and physical locations.

[0016] Based on the original phenotypic dataset, the minimum and maximum observed values ​​of the current population are independently calculated for each trait column. Then, the original data is mapped and transformed into a standardized phenotypic matrix with rows representing individuals and columns representing standardized trait values ​​through a linear transformation algorithm.

[0017] Individual quality control and site filtering are performed on the original gene data in the original genotype dataset to obtain qualified sites. The qualified sites are converted into a two-dimensional numerical matrix according to the genotype encoding rules and individual IDs and site coordinate indices are added to generate the genotype matrix.

[0018] In this scheme, the step of estimating genetic parameters and predicting breeding values ​​based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information specifically includes:

[0019] Obtain standardized phenotypic and genotype matrices. Based on the genotype matrix, use a weighted cosine similarity algorithm to calculate the allele similarity of whole-genome SNP sites among individuals, and generate a genome relationship matrix.

[0020] The standardized phenotypic matrix was used as the response variable, and the individual additive genetic effect was set as a random effect. A multi-trait animal model was constructed by combining the genomic relationship matrix, and the restricted maximum likelihood method was used for iterative solution.

[0021] Initialize the genetic and environmental variance parameters, update the variance component estimates iteratively using the expectation-maximization algorithm, calculate the conditional expectation of the observed data in each iteration, and maximize the likelihood function until the convergence threshold is reached, finally outputting the genetic parameter estimates.

[0022] By obtaining variance component estimates, a multi-trait optimal linear unbiased prediction algorithm is introduced. The genomic relationship matrix is ​​used as the covariance prior of random effects to calculate the genomic estimated breeding value of each trait for each individual, and breeding value prediction information is generated.

[0023] In this scheme, the step of analyzing the correlation between traits of the redfin snapper to be bred based on the breeding value prediction information and constructing a trait association network, and performing dynamic weight configuration for multiple traits to obtain a dynamic weight set, specifically includes:

[0024] The breeding value prediction information and breeding environment information are obtained. The maximum information coefficient algorithm is introduced to scan all trait combinations with the breeding value prediction information as input. The breeding value distribution interval of each trait is traversed in the form of a sliding window. The nonlinear association strength of different trait combinations at the genetic level is calculated, and the trait combination association strength matrix is ​​generated.

[0025] Subsequently, the acquired breeding environment information is discretized into several environmental intervals according to a preset division rule. The correlation strength is calculated independently and repeatedly in each environmental interval, and an environment-specific matrix is ​​generated based on the calculated correlation strength values.

[0026] The trait combination association strength matrix is ​​associated with the environment specificity matrix. Based on the matrix association result, an undirected weighted network is constructed with traits as nodes, the association strength between trait combinations as association weights, and the environment specificity matrix corresponding to the traits as auxiliary features.

[0027] Based on the undirected weighted network, the node centrality parameter and environmental sensitivity coefficient are extracted and input into the dynamic weight allocator to analyze the contribution weight of each trait to the comprehensive breeding value.

[0028] The dynamic weight allocator integrates the SHAP model and network constraint rules, calculates the conditional expectation Shapley value through individual whole genome markers and environmental parameters, and uses the calculated Shapley value to obtain the marginal contribution of each trait.

[0029] The dynamic weights of multiple traits are configured according to the marginal contribution of each trait to obtain the initial dynamic weights of multiple traits. The initial dynamic weights of multiple traits are then corrected by the network constraint rules to obtain the final dynamic weight set.

[0030] In this scheme, the step of calculating the genomic selection index and analyzing parental priority based on breeding value prediction information and dynamic weight set to generate a parental selection priority list specifically includes:

[0031] Obtain breeding value prediction information and dynamic weight set, generate a genome breeding value matrix based on the breeding value prediction information, perform standardization transformation on the genome breeding value matrix, and calculate the population mean and standard deviation of each trait column by column;

[0032] For each individual trait value, subtract the mean of the corresponding trait and divide by the standard deviation to generate a standardized breeding value matrix. The standardized breeding matrix is ​​then linearly weighted with the dynamic weight set to generate an initial genomic selection index.

[0033] Obtain the genotype matrix, extract the genotype codes of the target individuals in the preset QTL regions through the pre-trained QTL effect value model, multiply the genotype codes of each locus by the corresponding standardized effect value and then sum them to generate the QTL additive genetic contribution value.

[0034] The QTL additive genetic contribution value and the initial genomic selection index are arithmetically summed and fused to obtain the final genomic selection index. The parental priority of each candidate individual is generated based on the final genomic selection index.

[0035] A parent selection priority list is generated based on the parental priority corresponding to each candidate individual, sorted in descending order.

[0036] In this scheme, the process of generating a candidate parent pool through the parent selection priority list, performing mating combination analysis through the candidate parent pool, predicting genetic gain of the generated mating combinations, and generating a mating assistance scheme specifically includes:

[0037] Obtain a parent selection priority list, select several candidate parent individuals within a preset range according to the parent selection priority list to generate a candidate parent pool, extract a subset of the genotype matrix, calculate the kinship coefficient through the genome relationship matrix, and generate a kinship relationship edge weight matrix;

[0038] Based on the kinship edge weight matrix, a graph theory network is constructed with candidate parent individuals as nodes and kinship coefficients as edge weights. The graph theory network is then used as the input of the minimum spanning tree algorithm for breeding combination analysis.

[0039] Initialize an empty tree structure, start traversing from the node with the highest genomic selection index, and select the effective edge that connects to the current tree and has the smallest affinity coefficient each time. Limit the number of mating connections of each node by a preset pairing threshold, and calculate the inbreeding coefficient of the population after adding new edges to control inbreeding.

[0040] The iteration terminates when all nodes are connected or no legal edges can be added, and the minimum kinship cost mating tree is output. A mating list is generated based on the minimum kinship cost mating tree. For each mating combination in the mating list, the standardized genomic estimated breeding value of each trait of the parents is extracted, and the genotype distribution of the offspring is simulated through Mendel's laws of inheritance.

[0041] The simulation results of offspring genotype distribution are input into the QTL effect model to calculate the expected breeding value of offspring and predict genetic gain. Finally, mating combinations within the preset genetic gain range are selected to generate mating assistance schemes for recommendation.

[0042] A second aspect of the present invention provides a breeding assistance system for red snapper based on multi-trait synergistic selection. The system includes a memory and a processor. The memory contains a program for a red snapper breeding assistance method based on multi-trait synergistic selection. When executed by the processor, the program for the red snapper breeding assistance method based on multi-trait synergistic selection performs the following steps:

[0043] Phenotypic data and gene testing were conducted on the redfin snapper population to be bred, and a standardized phenotypic matrix and genotype matrix were constructed based on the collected data.

[0044] Genetic parameters are estimated and breeding values ​​are predicted based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information.

[0045] Based on the breeding value prediction information, the correlation between traits of the redfin snapper to be bred is analyzed and a trait association network is constructed. Then, the dynamic weights of multiple traits are configured to obtain a dynamic weight set.

[0046] Based on the breeding value prediction information and dynamic weight set, the genomic selection index is calculated and the parental priority is analyzed to generate a parental selection priority list.

[0047] A candidate parent pool is generated by the parent selection priority list, and mating combination analysis is performed on the candidate parent pool. Genetic gain prediction is then performed on the generated mating combinations to generate mating assistance schemes.

[0048] This invention discloses a breeding assistance method and system for red snapper based on multi-trait synergistic selection, comprising: collecting phenotypic data and performing gene detection to construct a standardized phenotypic matrix and genotype matrix; estimating genetic parameters and predicting breeding values ​​based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information; analyzing the association relationships between traits of the red snapper to be bred and constructing a trait association network, performing dynamic weight allocation for multiple traits to obtain a dynamic weight set; calculating the genomic selection index and analyzing parental priority based on the breeding value prediction information and the dynamic weight set to generate a parental selection priority list; generating a candidate parent pool through the parental selection priority list, performing mating combination analysis, and predicting the genetic gain of the generated mating combinations to generate a mating assistance scheme. This improves the accuracy and balance of genetic gain in red snapper breeding, overcomes the limitations of traditional empirical mating, and improves breeding efficiency. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments or examples of the present invention, the drawings used in the embodiments or examples will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained according to these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a breeding assistance method for redfin snapper based on multi-trait synergistic selection, provided in an embodiment of the present invention.

[0051] Figure 2 This is a block diagram of a breeding assistance system for redfin snapper based on multi-trait synergistic selection, provided in an embodiment of the present invention.

[0052] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0055] Figure 1 This is a flowchart of a breeding assistance method for redfin snapper based on multi-trait synergistic selection, provided in an embodiment of the present invention.

[0056] like Figure 1 As shown, the present invention provides a flowchart of a breeding assistance method for redfin snapper based on multi-trait synergistic selection, including:

[0057] S102, Phenotypic data and gene testing were conducted on the redfin snapper population to be bred, and a standardized phenotypic matrix and genotype matrix were constructed based on the collected data;

[0058] S104, Genetic parameters are estimated and breeding values ​​are predicted based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information;

[0059] S106, Based on the breeding value prediction information, analyze the correlation between traits of the redfin snapper to be bred and construct a trait association network, and perform dynamic weight configuration for multiple traits to obtain a dynamic weight set;

[0060] S108: Based on the breeding value prediction information and dynamic weight set, the genomic selection index is calculated and the parental priority is analyzed to generate a parental selection priority list;

[0061] S110, a candidate parent pool is generated through the parent selection priority list, mating combination analysis is performed through the candidate parent pool, and genetic gain prediction is performed on the generated mating combinations to generate a mating assistance scheme.

[0062] Furthermore, in a preferred embodiment of the present invention, the step of collecting phenotypic data and performing gene detection on the redfin snapper population to be bred, and constructing a standardized phenotypic matrix and genotype matrix based on the collected data, specifically includes:

[0063] The weighing platform was used to weigh the red snapper population to be bred in the target breeding pond within a preset collection period. The body length growth rate of the red snapper population to be bred within the preset collection period was calculated by deploying a multi-source sensor array and combining machine vision technology.

[0064] The amount of uneaten feed in the target breeding pond is obtained by deploying uneaten feed collection devices. The difference between the uneaten feed and the daily feed amount is calculated to obtain the feed intake data. The feed intake-weight gain ratio is calculated using the obtained fish body weight data.

[0065] Based on a pre-set disease resistance breeding test plan, a disease resistance test was conducted on redfin snapper in the target breeding pond. The survival rate within the pre-set time period was recorded as the quantitative value of the disease resistance trait. Finally, the original phenotypic dataset was obtained by combining the fish body identification.

[0066] Subsequently, gene samples were collected from the redfin snapper population to be bred. Based on the collected gene samples, whole-genome typing was performed using gene detection technology to form an original genotype dataset containing locus genotypes and physical locations.

[0067] Based on the original phenotypic dataset, the minimum and maximum observed values ​​of the current population are independently calculated for each trait column. Then, the original data is mapped and transformed into a standardized phenotypic matrix with rows representing individuals and columns representing standardized trait values ​​through a linear transformation algorithm.

[0068] Individual quality control and site filtering are performed on the original gene data in the original genotype dataset to obtain qualified sites. The qualified sites are converted into a two-dimensional numerical matrix according to the genotype encoding rules and individual IDs and site coordinate indices are added to generate the genotype matrix.

[0069] It should be noted that the integrated dynamic weighing platform is used to weigh the redfin snapper population in the target breeding pond at a preset collection cycle (e.g., every 7 days). Simultaneously, an underwater multi-source sensor array (including depth cameras and laser scanners) combined with machine vision algorithms tracks the individual body length changes, and the body length growth rate of each fish within the cycle is calculated based on time series analysis. Uneaten feed is collected daily through a bottom siphon pipe using a deployed uneaten feed collection device. After drying and weighing, the uneaten feed is compared with the amount of feed recorded by the automatic feeding system to obtain the precise feed intake. This is then combined with weight gain data to calculate the feed conversion efficiency (feed intake - weight gain ratio). Disease resistance quantification follows a standardized pathogen challenge test procedure: randomized grouping for testing, recording survival status during a 96-hour observation period, and calculating individual survival rate as the disease resistance phenotypic value. The phenotypic data is then associated with fish identification markers (e.g., PIT tag encoding) to form an original phenotypic dataset covering the original observations of three key traits: growth, feed efficiency, and disease resistance. Simultaneously, fin samples from a population of redfin snapper were collected for whole-genome SNP genotyping, generating an original genotype dataset containing genotypes (AA / AB / BB) and physical location information for several loci. Individual-level quality control was then performed, removing invalid individuals with a detection rate below 95%. Locus-level filtering was also conducted to remove low-quality loci with a detection rate below 98% or a minor allele frequency below 0.05. Finally, qualified loci were uniformly represented as a 0 / 1 / 2 numerical matrix (AA=0, AB=1, BB=2) through encoding transformation. After adding individual ID indexes and SNP physical location annotations, a structured genotype matrix was formed. Finally, each trait column in the original phenotypic dataset is processed independently: the minimum and maximum observed values ​​of the entire population for body length growth rate, feed conversion rate, and disease resistance survival rate are extracted, and the original data are mapped to the [0,1] interval using a linear transformation formula. For example, the normalized value of the growth trait = (actual weight gain - minimum weight gain) / (maximum weight gain - minimum weight gain). The resulting standardized phenotypic matrix has rows corresponding to individual IDs and columns carrying dimensionless trait values, providing unbiased input for subsequent genetic evaluation. The genotype matrix retains the effective site information of genetic variation through quality control filtering and encoding transformation, and shares the individual ID index system with the phenotypic matrix, together forming a standardized data base for breeding analysis.

[0070] Furthermore, in a preferred embodiment of the present invention, the step of estimating genetic parameters and predicting breeding values ​​based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information specifically includes:

[0071] Obtain standardized phenotypic and genotype matrices. Based on the genotype matrix, use a weighted cosine similarity algorithm to calculate the allele similarity of whole-genome SNP sites among individuals, and generate a genome relationship matrix.

[0072] The standardized phenotypic matrix was used as the response variable, and the individual additive genetic effect was set as a random effect. A multi-trait animal model was constructed by combining the genomic relationship matrix, and the restricted maximum likelihood method was used for iterative solution.

[0073] Initialize the genetic and environmental variance parameters, update the variance component estimates iteratively using the expectation-maximization algorithm, calculate the conditional expectation of the observed data in each iteration, and maximize the likelihood function until the convergence threshold is reached, finally outputting the genetic parameter estimates.

[0074] By obtaining variance component estimates, a multi-trait optimal linear unbiased prediction algorithm is introduced. The genomic relationship matrix is ​​used as the covariance prior of random effects to calculate the genomic estimated breeding value of each trait for each individual, and breeding value prediction information is generated.

[0075] It should be noted that, firstly, a weighted cosine similarity algorithm is used to calculate the genome-wide genetic similarity between individuals. By assigning weights to each SNP locus (based on allele frequency to calculate information weights), the degree of allele matching among individuals at tens of thousands of loci is quantified using the cosine similarity formula, ultimately generating a symmetric genome relation matrix (GRM), whose element values ​​characterize the molecular kinship strength between individuals (ranging from -1 to 1, with positive values ​​indicating genetic similarity). Subsequently, a multi-trait animal model is constructed for genetic parameter estimation: using the standardized phenotypic matrix as the response variable, individual additive genetic effects are set as random effects, and their covariance structure is defined by the genome relation matrix. A restricted maximum likelihood method is used to initiate iterative solution: the genetic variance and environmental variance components are initialized, and the parameters are updated iteratively using the expectation-maximization algorithm. In each iteration, the conditional expectation of the observed data is calculated, and the restricted likelihood function is maximized until the change in the genetic variance estimate is less than a preset threshold, at which point convergence is determined, and key genetic parameter information is output, including the heritability of each trait, the genetic correlation coefficient between traits, and the estimated variance components. Based on the convergent variance component results, a multi-trait optimal linear unbiased prediction algorithm is introduced to calculate the genomic breeding value. The genomic relationship matrix is ​​used as the covariance prior for random effects, and a mixture model equation system is constructed: the fixed effects part includes the environmental factor design matrix, and the random effects part is constrained by the GRM for its covariance structure. A sparse matrix solver is used to calculate the genomic estimated breeding value (GEBV) for each individual on each trait. The final generated breeding value prediction information forms a structured data table (row = individual ID, column = three-trait GEBV value). This table also carries an accuracy index for genetic parameter estimation, providing a quantitative basis for subsequent selection decisions regarding genetic potential.

[0076] Furthermore, in a preferred embodiment of the present invention, the step of analyzing the correlation between traits of the redfin snapper to be bred based on the breeding value prediction information and constructing a trait association network, and performing dynamic weight configuration for multiple traits to obtain a dynamic weight set, specifically includes:

[0077] The breeding value prediction information and breeding environment information are obtained. The maximum information coefficient algorithm is introduced to scan all trait combinations with the breeding value prediction information as input. The breeding value distribution interval of each trait is traversed in the form of a sliding window. The nonlinear association strength of different trait combinations at the genetic level is calculated, and the trait combination association strength matrix is ​​generated.

[0078] Subsequently, the acquired breeding environment information is discretized into several environmental intervals according to a preset division rule. The correlation strength is calculated independently and repeatedly in each environmental interval, and an environment-specific matrix is ​​generated based on the calculated correlation strength values.

[0079] The trait combination association strength matrix is ​​associated with the environment specificity matrix. Based on the matrix association result, an undirected weighted network is constructed with traits as nodes, the association strength between trait combinations as association weights, and the environment specificity matrix corresponding to the traits as auxiliary features.

[0080] Based on the undirected weighted network, the node centrality parameter and environmental sensitivity coefficient are extracted and input into the dynamic weight allocator to analyze the contribution weight of each trait to the comprehensive breeding value.

[0081] The dynamic weight allocator integrates the SHAP model and network constraint rules, calculates the conditional expectation Shapley value through individual whole genome markers and environmental parameters, and uses the calculated Shapley value to obtain the marginal contribution of each trait.

[0082] The dynamic weights of multiple traits are configured according to the marginal contribution of each trait to obtain the initial dynamic weights of multiple traits. The initial dynamic weights of multiple traits are then corrected by the network constraint rules to obtain the final dynamic weight set.

[0083] It should be noted that the Genomic Breeding Value (GEBV) prediction information and time-series data of environmental parameters (water temperature, salinity, dissolved oxygen) are obtained, and the Maximum Information Coefficient (MIC) algorithm is introduced for multi-dimensional association analysis. Using the GEBV matrix as input, a sliding window mechanism is used to scan all trait combinations, calculating the nonlinear association strength at the genetic level: the window slides along the trait value distribution interval, the mutual information entropy of trait pairs within each bin is calculated, and after normalization, the MIC value (0-1 interval, larger values ​​indicate stronger association) is output, generating a symmetric trait combination association strength matrix. Environmental data is processed simultaneously: continuous environmental parameters are discretized into mutually exclusive intervals based on a preset threshold. The MIC calculation is independently repeated within each environmental interval subset to generate an environment-specific association matrix. The genetic-level association matrix and the environmental matrix are spatiotemporally aligned, and an undirected weighted network is constructed based on the alignment result, with traits as nodes (node ​​size proportional to cross-environment association stability), MIC values ​​as edge weights, and the corresponding environment-specific matrix as an auxiliary feature. Subsequently, key parameters are extracted using an undirected weighted network: centrality is calculated using node weighting, and sensitivity coefficients are fitted using edge weight environmental gradients. The centrality and sensitivity coefficients are input into a dynamic weight allocator, which integrates a dual-channel processing engine: a SHAP parsing channel, which uses individual whole-genome SNP data and real-time environmental parameters as input to calculate the marginal contribution of each trait using the conditional expectation Shapley value; and a network constraint channel, which applies centrality to set a lower limit for weights (centrality > 0.8 forces weights ≥ 0.3) and dynamically modulates the contribution using the sensitivity coefficient (e.g., contribution of disease resistance under high temperature × 1.5). The initial dynamic weights for multiple traits are corrected using the aforementioned network constraint rules. Finally, a dynamic weight set for the environmental index is generated.

[0084] Furthermore, in a preferred embodiment of the present invention, the step of calculating the genomic selection index and analyzing parental priority based on breeding value prediction information and dynamic weight set to generate a parental selection priority list specifically includes:

[0085] Obtain breeding value prediction information and dynamic weight set, generate a genome breeding value matrix based on the breeding value prediction information, perform standardization transformation on the genome breeding value matrix, and calculate the population mean and standard deviation of each trait column by column;

[0086] For each individual trait value, subtract the mean of the corresponding trait and divide by the standard deviation to generate a standardized breeding value matrix. The standardized breeding matrix is ​​then linearly weighted with the dynamic weight set to generate an initial genomic selection index.

[0087] Obtain the genotype matrix, extract the genotype codes of the target individuals in the preset QTL regions through the pre-trained QTL effect value model, multiply the genotype codes of each locus by the corresponding standardized effect value and then sum them to generate the QTL additive genetic contribution value.

[0088] The QTL additive genetic contribution value and the initial genomic selection index are arithmetically summed and fused to obtain the final genomic selection index. The parental priority of each candidate individual is generated based on the final genomic selection index.

[0089] A parent selection priority list is generated based on the parental priority corresponding to each candidate individual, sorted in descending order.

[0090] It should be noted that, firstly, a structured genomic breeding value matrix is ​​constructed: rows correspond to individual IDs, and columns store the original breeding values ​​for the three traits of growth rate, disease resistance, and feed efficiency, respectively. This matrix is ​​then standardized, and the population mean and standard deviation for each trait are calculated column by column. For each individual, the trait value is subtracted from the corresponding mean and divided by the standard deviation to generate a Z-score standardized matrix. This eliminates dimensional differences between traits and ensures the fairness of index synthesis. Next, the standardized breeding value matrix is ​​linearly weighted and fused with a dynamic weight set: the standardized values ​​of the three traits for each individual are multiplied by the dynamic weight set under the current environment, and the products are accumulated to generate an initial genomic selection index, forming an initial index list carrying individual IDs and environmental labels, representing the comprehensive genetic potential under dynamically adjusted weights. Simultaneously, a pre-trained QTL effect value model is invoked to extract the genotype codes of the target individuals in the preset core QTL regions from the genotype matrix. Effect value transformation is performed on each locus, and the genotype code is multiplied by the corresponding standardized effect value, accumulating to generate the QTL additive genetic contribution value. Subsequently, the additive genetic contribution value of the QTL is arithmetically summed with the initial genomic selection index to generate the final genomic selection index. Based on this index value, the parental priority of candidate individuals is generated (the larger the value, the higher the priority), and a list of preferred parental priorities is output in descending order.

[0091] Furthermore, in a preferred embodiment of the present invention, the step of generating a candidate parent pool through the parent selection priority list, performing mating combination analysis through the candidate parent pool, predicting genetic gain of the generated mating combinations, and generating a mating assistance scheme specifically includes:

[0092] Obtain a parent selection priority list, select several candidate parent individuals within a preset range according to the parent selection priority list to generate a candidate parent pool, extract a subset of the genotype matrix, calculate the kinship coefficient through the genome relationship matrix, and generate a kinship relationship edge weight matrix;

[0093] Based on the kinship edge weight matrix, a graph theory network is constructed with candidate parent individuals as nodes and kinship coefficients as edge weights. The graph theory network is then used as the input of the minimum spanning tree algorithm for breeding combination analysis.

[0094] Initialize an empty tree structure, start traversing from the node with the highest genomic selection index, and select the effective edge that connects to the current tree and has the smallest kinship coefficient each time. Limit the number of mating connections of each node by a preset pairing threshold, and calculate the inbreeding coefficient of the population after adding new edges to control inbreeding.

[0095] The iteration terminates when all nodes are connected or no legal edges can be added, and the minimum kinship cost mating tree is output. A mating list is generated based on the minimum kinship cost mating tree. For each mating combination in the mating list, the standardized genomic estimated breeding value of each trait of the parents is extracted, and the genotype distribution of the offspring is simulated through Mendel's laws of inheritance.

[0096] The simulation results of offspring genotype distribution are input into the QTL effect model to calculate the expected breeding value of offspring and predict genetic gain. Finally, mating combinations within the preset genetic gain range are selected to generate mating assistance schemes for recommendation.

[0097] It should be noted that, after obtaining the parent selection priority list, candidate individuals ranked within a preset range (e.g., the top 20%) are selected to form a candidate parent pool, and a subset of the genotype matrix of the corresponding individuals is extracted simultaneously. The kinship coefficient between each pair of individuals is calculated using the genomic relationship matrix (the formula is the standardized value of the shared allele frequency), generating a symmetric kinship edge weight matrix (matrix element values ​​range from 0 to 1, with smaller values ​​indicating more distant kinship). Subsequently, a graph theory network is constructed based on this edge weight matrix: candidate parent individuals are used as nodes (node ​​size maps to their genomic selection index), and kinship coefficients are used as edge weights (only edges with coefficients <0.3 are retained to avoid the risk of inbreeding). This network is then input into the minimum spanning tree algorithm: an empty tree structure is initialized, and the node with the highest GSI is selected as the starting root node. During iterative expansion, priority is given to selecting the effective edge connected to the current tree with the smallest kinship coefficient (i.e., the most distant legal pairing). Each time a new edge is added, the number of connections between nodes is limited by a preset pairing threshold (e.g., ≤2 times). The average inbreeding coefficient ΔF of the population after adding the new edge is calculated in real time (calculation formula: ΔF = 0.5 × kinship coefficient of the newly added edge), ensuring that ΔF is always <0.01. The iteration terminates when all nodes are connected or there are no more legal edges to add, and the minimum kinship cost mating tree is output (tree edges represent the optimal mating combination). Next, a mating list is generated based on the mating tree structure. For each mating combination in the list, the standardized genome estimates of the parents' breeding values ​​(Z-scores of growth / disease resistance / feed efficiency) are extracted. The offspring genotype distribution is simulated using Mendelian inheritance laws: allele segregation is calculated independently for each SNP locus (e.g., when the father is AB and the mother is AA, the offspring genotype probability is 50% AA + 50% AB). The offspring genotype probability distribution is input into a pre-trained QTL effect model to calculate the expected breeding value of the offspring. Specifically, for each locus in the target QTL interval, the probabilities of offspring genotypes are multiplied by standardized effect values ​​and then summed to generate predicted breeding values ​​for the three traits. Genetic gain (ΔG = offspring mean - parent mean) is calculated based on the difference between the average breeding values ​​of the offspring and parent populations. Finally, mating combinations with ΔG exceeding a preset threshold are selected, generating mating assistance plans (including a list of recommended mating pairs, a genetic gain heatmap, and inbreeding risk rating) to guide the mating schedule and family establishment plan of broodstock, achieving precise control over genetic progress.

[0098] Figure 2 A breeding assistance system 2 for red snapper based on multi-trait synergistic selection is provided in one embodiment of the present invention. The system includes: a memory 21 and a processor 22. The memory 21 contains a breeding assistance method program for red snapper based on multi-trait synergistic selection. When the processor 22 executes the breeding assistance method program for red snapper based on multi-trait synergistic selection, it performs the following steps:

[0099] Phenotypic data and gene testing were conducted on the redfin snapper population to be bred, and a standardized phenotypic matrix and genotype matrix were constructed based on the collected data.

[0100] Genetic parameters are estimated and breeding values ​​are predicted based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information.

[0101] Based on the breeding value prediction information, the correlation between traits of the redfin snapper to be bred is analyzed and a trait association network is constructed. Then, the dynamic weights of multiple traits are configured to obtain a dynamic weight set.

[0102] Based on the breeding value prediction information and dynamic weight set, the genomic selection index is calculated and the parental priority is analyzed to generate a parental selection priority list.

[0103] A candidate parent pool is generated by the parent selection priority list, and mating combination analysis is performed on the candidate parent pool. Genetic gain prediction is then performed on the generated mating combinations to generate mating assistance schemes.

[0104] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0105] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0106] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0107] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0109] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A breeding assistance method for redfin snapper based on multi-trait synergistic selection, characterized in that, include: Phenotypic data and gene testing were conducted on the redfin snapper population to be bred, and a standardized phenotypic matrix and genotype matrix were constructed based on the collected data. Genetic parameters are estimated and breeding values ​​are predicted based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information. Based on the breeding value prediction information, the correlation between traits of the redfin snapper to be bred is analyzed and a trait association network is constructed. Then, the dynamic weights of multiple traits are configured to obtain a dynamic weight set. Based on the breeding value prediction information and dynamic weight set, the genomic selection index is calculated and the parental priority is analyzed to generate a parental selection priority list. A candidate parent pool is generated from the parent selection priority list, and mating combination analysis is performed on the candidate parent pool. Genetic gain prediction is then performed on the generated mating combinations to generate mating assistance schemes. The process of collecting phenotypic data and performing gene testing on the redfin snapper population to be bred, and constructing a standardized phenotypic matrix and genotype matrix based on the collected data, specifically includes: The weighing platform was used to weigh the red snapper population to be bred in the target breeding pond within a preset collection period. The body length growth rate of the red snapper population to be bred within the preset collection period was calculated by deploying a multi-source sensor array and combining machine vision technology. The amount of uneaten feed in the target breeding pond is obtained by deploying uneaten feed collection devices. The difference between the uneaten feed and the daily feed amount is calculated to obtain the feed intake data. The feed intake-weight gain ratio is calculated using the obtained fish body weight data. Based on a pre-set disease resistance breeding test plan, a disease resistance test was conducted on redfin snapper in the target breeding pond. The survival rate within the pre-set time period was recorded as the quantitative value of the disease resistance trait. Finally, the original phenotypic dataset was obtained by combining the fish body identification. Subsequently, gene samples were collected from the redfin snapper population to be bred. Based on the collected gene samples, whole-genome typing was performed using gene detection technology to form an original genotype dataset containing locus genotypes and physical locations. Based on the original phenotypic dataset, the minimum and maximum observed values ​​of the current population are independently calculated for each trait column. Then, the original data is mapped and transformed into a standardized phenotypic matrix with rows representing individuals and columns representing standardized trait values ​​through a linear transformation algorithm. Individual quality control and site filtering are performed on the original gene data in the original genotype dataset to obtain qualified sites. The qualified sites are converted into a two-dimensional numerical matrix according to the genotype encoding rules and individual IDs and site coordinate indices are added to generate the genotype matrix.

2. The breeding assistance method for redfin snapper based on multi-trait synergistic selection according to claim 1, characterized in that, The process of estimating genetic parameters and predicting breeding values ​​based on the standardized phenotypic matrix and genotype matrix to obtain breeding value prediction information specifically includes: Obtain standardized phenotypic and genotype matrices. Based on the genotype matrix, use a weighted cosine similarity algorithm to calculate the allele similarity of whole-genome SNP sites among individuals, and generate a genome relationship matrix. The standardized phenotypic matrix was used as the response variable, and the individual additive genetic effect was set as a random effect. A multi-trait animal model was constructed by combining the genomic relationship matrix, and the restricted maximum likelihood method was used for iterative solution. Initialize the genetic and environmental variance parameters, update the variance component estimates iteratively using the expectation-maximization algorithm, calculate the conditional expectation of the observed data in each iteration, and maximize the likelihood function until the convergence threshold is reached, finally outputting the genetic parameter estimates. By obtaining variance component estimates, a multi-trait optimal linear unbiased prediction algorithm is introduced. The genomic relationship matrix is ​​used as the covariance prior of random effects to calculate the genomic estimated breeding value of each trait for each individual, and breeding value prediction information is generated.

3. The breeding assistance method for redfin snapper based on multi-trait synergistic selection according to claim 1, characterized in that, The process involves analyzing the correlation between traits of the redfin snapper to be bred based on the breeding value prediction information, constructing a trait association network, and configuring dynamic weights for multiple traits to obtain a dynamic weight set. Specifically, this includes: The breeding value prediction information and breeding environment information are obtained. The maximum information coefficient algorithm is introduced to scan all trait combinations with the breeding value prediction information as input. The breeding value distribution interval of each trait is traversed in the form of a sliding window. The nonlinear association strength of different trait combinations at the genetic level is calculated, and the trait combination association strength matrix is ​​generated. Subsequently, the acquired breeding environment information is discretized into several environmental intervals according to a preset division rule. The correlation strength is calculated independently and repeatedly in each environmental interval, and an environment-specific matrix is ​​generated based on the calculated correlation strength values. The trait combination association strength matrix is ​​associated with the environment specificity matrix. Based on the matrix association result, an undirected weighted network is constructed with traits as nodes, the association strength between trait combinations as association weights, and the environment specificity matrix corresponding to the traits as auxiliary features. Based on the undirected weighted network, the node centrality parameter and environmental sensitivity coefficient are extracted and input into the dynamic weight allocator to analyze the contribution weight of each trait to the comprehensive breeding value. The dynamic weight allocator integrates the SHAP model and network constraint rules, calculates the conditional expectation Shapley value through individual whole genome markers and environmental parameters, and uses the calculated Shapley value to obtain the marginal contribution of each trait. The dynamic weights of multiple traits are configured according to the marginal contribution of each trait to obtain the initial dynamic weights of multiple traits. The initial dynamic weights of multiple traits are then corrected by the network constraint rules to obtain the final dynamic weight set.

4. The breeding assistance method for redfin snapper based on multi-trait synergistic selection according to claim 1, characterized in that, The step of calculating the genomic selection index and analyzing parental priority based on breeding value prediction information and dynamic weight set to generate a parental selection priority list specifically includes: Obtain breeding value prediction information and dynamic weight set, generate a genome breeding value matrix based on the breeding value prediction information, perform standardization transformation on the genome breeding value matrix, and calculate the population mean and standard deviation of each trait column by column; For each individual trait value, subtract the mean of the corresponding trait and divide by the standard deviation to generate a standardized breeding value matrix. Linearly weight the standardized breeding value matrix with the dynamic weight set to generate an initial genomic selection index. Obtain the genotype matrix, extract the genotype codes of the target individuals in the preset QTL regions through the pre-trained QTL effect value model, multiply the genotype codes of each locus by the corresponding standardized effect value and then sum them to generate the QTL additive genetic contribution value. The QTL additive genetic contribution value and the initial genomic selection index are arithmetically summed and fused to obtain the final genomic selection index. The parental priority of each candidate individual is generated based on the final genomic selection index. A parent selection priority list is generated based on the parental priority corresponding to each candidate individual, sorted in descending order.

5. The breeding assistance method for redfin snapper based on multi-trait synergistic selection according to claim 1, characterized in that, The process of generating a candidate parent pool from the parent selection priority list, performing mating combination analysis on the candidate parent pool, predicting genetic gain from the generated mating combinations, and generating a mating assistance plan specifically includes: Obtain a parent selection priority list, select several candidate parent individuals within a preset range according to the parent selection priority list to generate a candidate parent pool, extract a subset of the genotype matrix, calculate the kinship coefficient through the genome relationship matrix, and generate a kinship relationship edge weight matrix; Based on the kinship edge weight matrix, a graph theory network is constructed with candidate parent individuals as nodes and kinship coefficients as edge weights. The graph theory network is then used as the input of the minimum spanning tree algorithm for breeding combination analysis. Initialize an empty tree structure, start traversing from the node with the highest genomic selection index, and select the effective edge that connects to the current tree and has the smallest affinity coefficient each time. Limit the number of mating connections of each node by a preset pairing threshold, and calculate the inbreeding coefficient of the population after adding new edges to control inbreeding. The iteration terminates when all nodes are connected or no legal edges can be added, and the minimum kinship cost mating tree is output. A mating list is generated based on the minimum kinship cost mating tree. For each mating combination in the mating list, the standardized genomic estimated breeding value of each trait of the parents is extracted, and the genotype distribution of the offspring is simulated through Mendel's laws of inheritance. The simulation results of offspring genotype distribution are input into the QTL effect model to calculate the expected breeding value of offspring and predict genetic gain. Finally, mating combinations within the preset genetic gain range are selected to generate mating assistance schemes for recommendation.

6. A breeding assistance system for redfin snapper based on multi-trait synergistic selection, characterized in that, The system includes: a memory and a processor. The memory contains a program for a breeding assistance method for red snapper based on multi-trait synergistic selection. When the program for the breeding assistance method for red snapper based on multi-trait synergistic selection is executed by the processor, it implements the breeding assistance method for red snapper based on multi-trait synergistic selection as described in any one of claims 1-5.

Images