A Hybridization and Optimization Method for High-Iron Strains in Auricularia auricula-judae Based on Big Data Analysis

Abstract

This invention provides a method for optimizing the hybridization and selection of high-iron strains in black fungus based on big data analysis. It relates to the fields of big data analysis and black fungus variety breeding. The method includes: Step S1: Collecting basic data of strains, hybridization process data, culture environment parameters, and external reference data, and performing standardization and structuring processing; Step S2: Encoding and optimizing hybridization combinations using a biased random bond genetic algorithm, combined with a dual-elite evolutionary mechanism and dynamic parameter adjustment strategy to generate the optimal hybridization scheme; Step S3: Constructing a dynamic environmental model using a model-based hybrid inverse reinforcement learning algorithm, searching for the optimal culture strategy that maximizes iron content in a virtual environment; Step S4: Constructing a multi-index comprehensive scoring system based on the analytic hierarchy process (AHP), identifying abnormal samples and triggering model backtracking adjustments, outputting high-performance strains for the next round of breeding. This invention significantly improves breeding efficiency and the accuracy of strain selection.

Description

Technical Field

[0001] This invention relates to the fields of big data analysis and black fungus variety breeding technology, and in particular to a method for hybridization and optimization of high-iron strains of black fungus based on big data analysis. Background Technology

[0002] Black fungus, a nutrient-rich edible and medicinal fungus, is significant for improving anemia and supplementing trace elements due to its high iron content. Currently, there are few reports on the breeding of functional varieties of black fungus, while research has been conducted on other edible and medicinal fungi such as Ganoderma lucidum, Cordyceps militaris, and Hericium erinaceus, resulting in the breeding of functional varieties, such as polysaccharide-rich Ganoderma lucidum, cordycepin-rich Cordyceps militaris, and polysaccharide-rich Hericium erinaceus. These functional variety breeding methods often employ mutation breeding or combine it with hybridization breeding and systematic selection breeding, but these methods suffer from problems such as non-directionality, randomness, unstable phenotypic expression, and high workload. Furthermore, traditional strain hybridization breeding generally faces the following problems: low combination screening efficiency: traditional hybridization usually uses a full combination approach, resulting in a large number of combinations, low success rate, and a lack of effective screening and process simulation mechanisms. Long breeding cycle and high cost: extensive field experiments are required to verify target traits, such as iron content, yield, and growth period. Hybrid combination evaluation relies on manual experience: in the strain performance evaluation stage, there is often a lack of systematic quantitative standards, leading to inaccurate identification of superior strains. Sensitive to the culture environment but lacking intelligent prediction methods: traits such as iron content are significantly affected by factors such as culture medium formulation, temperature and humidity, and traditional methods cannot dynamically predict and optimize them.

[0003] Although big data and artificial intelligence technologies have been applied in crop breeding, they still have limitations in the field of edible and medicinal fungi: existing genetic algorithms are difficult to adapt to the high-dimensional combination optimization problem of strain hybridization, and premature convergence leads to insufficient global search capabilities; inverse reinforcement learning technology is mostly focused on fields such as robot control and has not yet been used for modeling culture conditions, and lacks a hybrid driving mechanism of expert data and real-time data; dynamic environment-gene expression coupling models are lacking, making it impossible to predict the probabilistic impact of gene recombination on iron content.

[0004] Therefore, it is urgent to construct a set of intelligent breeding methods for high-iron strains of black fungus that integrates combinatorial optimization algorithms, environmental dynamic modeling, and multi-objective evaluation, so as to achieve closed-loop optimization of hybridization design, culture regulation, and performance feedback. Summary of the Invention

[0005] To address the aforementioned issues, this invention aims to construct a method for hybridization and optimization of high-iron strains of black fungus based on big data analysis, forming a closed-loop automated optimization process from data collection, combination optimization, environmental prediction to performance evaluation.

[0006] To achieve the above objectives, the following technical solution is adopted:

[0007] In a first aspect, the present invention provides a method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis, comprising:

[0008] Step S1: Collect basic data of the strain, hybridization process data, culture environment parameters and external reference data, and perform standardization and structuring processing;

[0009] Step S2: Use a biased random bond genetic algorithm to encode and optimize the hybridization combination, and combine a dual-elite evolutionary mechanism and a dynamic parameter adjustment strategy to generate the optimal hybridization scheme;

[0010] Step S3: Construct a dynamic environmental model using a hybrid inverse reinforcement learning algorithm with a model, and search for the optimal cultivation strategy that maximizes iron content in the virtual environment;

[0011] Step S4: Construct a multi-index comprehensive scoring system based on the analytic hierarchy process (AHP) to identify abnormal samples and trigger model backtracking adjustments, and output high-performance strains to enter the next round of breeding.

[0012] Furthermore, the implementation of the biased random key genetic algorithm in step S2 includes:

[0013] The hybridization combination is encoded as a random key chromosome vector X represented by n-dimensional continuous real numbers;

[0014] The random bond chromosome vector X is decoded into a specific combination strategy and fitness score by an encoder. The objective function is constructed with iron content over-parent rate and agronomic trait stability as the evaluation core.

[0015] By introducing parental gene expression data, simulating gene recombination behavior during the fusion of different monokaryotic strains, predicting the probability distribution of iron content in offspring, and forming a dynamic hybridization simulation model;

[0016] Employing a dual-elite evolutionary mechanism, each generation retains the previous generation's fitness. There are elite individuals, and a biased crossover operation is performed with probability p>0.5: the offspring genes inherit the elite parent 1 gene with probability p, and inherit the non-elite parent 2 gene with probability 1-p.

[0017] As the number of iterations progresses, parameters such as the elite ratio and mutation probability are dynamically adjusted to adapt to the search progress.

[0018] Furthermore, the biased random key genetic algorithm further includes:

[0019] Jittering mechanism: applying probability to elite individuals Apply a disturbance;

[0020] Reset mechanism: When there is no improvement in fitness after a series of preset generations, remove non-elite individuals and regenerate random chromosomes.

[0021] Furthermore, the objective function is:

[0022]

[0023] in, : Fitness score corresponding to chromosome X; Iron content exceeding parental rate, which is the percentage of offspring predicted to have iron content higher than the average of the parents; : Stability index of agronomic traits; α, β: Weight parameters of the objective function, set according to task requirements.

[0024] Furthermore, step S3: constructing a dynamic environmental model using a hybrid inverse reinforcement learning algorithm, and searching for the optimal cultivation strategy to maximize iron content in a virtual environment, includes:

[0025] A hybrid dataset was constructed using expert demonstration data and online learner data.

[0026] A neural network is used to fit the environmental state transition probabilities to obtain an environmental dynamic model. Furthermore, a closed-loop learning mechanism is formed by strategy optimization and reward function iteration to efficiently simulate the adjustment behavior of culture medium formulation and environmental conditions in a virtual model environment.

[0027] Furthermore, step S3 also includes:

[0028] The states are randomly selected from the expert demonstration dataset as the initial states for policy optimization.

[0029] In environmental dynamics model Within this framework, a gradient-free optimization algorithm or model predictive control method is used to optimize policy π to maximize the cumulative reward, and the reward function is updated through gradient ascent to generate a new policy π. This data is further fed back into the real environment, generating new data. This forms a closed data loop.

[0030] Furthermore, the iterative formula for the reward function is:

[0031]

[0032] The current reward function, used in the t-th iteration; The reward function updated after the (t+1)th iteration; The learning rate is used to control the update step size of the reward function. : Calculate the gradient of the reward function f; In the reward function Below, expert strategy Expected cumulative reward; In the reward function Below, the current strategy Expected cumulative reward; Expert strategies typically stem from human experience or high-quality demonstrations. : The current strategy, i.e. the strategy optimized in the t-th iteration.

[0033] Furthermore, step S4: constructing a multi-index comprehensive scoring system based on the analytic hierarchy process (AHP), identifying abnormal samples and triggering model backtracking adjustments, and outputting high-performance strains for the next round of breeding, includes:

[0034] The Analytic Hierarchy Process (AHP) was used to assign weights to each indicator: iron content 50%, yield 30%, and growth period 20%. The comprehensive score for each strain was calculated as follows:

[0035] Overall score = 0.5 × iron content score + 0.3 × yield score + 0.2 × reproductive period score

[0036] Each score is standardized.

[0037] When the overall score of the strain is lower than the preset threshold or the iron content is not up to standard, the system automatically backtracks to the parent selection or hybridization design stage and updates the parameters of the genetic algorithm or recommendation model to optimize the prediction quality of mating combinations.

[0038] Furthermore, step S4 also includes:

[0039] If the strain compliance rate is lower than the threshold, the weight of the high-iron trait is increased or a penalty term is added. The total loss is calculated as follows:

[0040] Total loss = Original loss + Penalty coefficient × (Average iron content of parents - Actual iron content) 2

[0041] The penalty coefficient is dynamically adjusted based on the degree to which the iron content fails to meet the standard.

[0042] Secondly, this invention also provides a hybridization and optimization system for high-iron strains of Auricularia auricula-judae based on big data analysis, comprising:

[0043] Data acquisition and preprocessing module: Collects basic data of strains, hybridization process data, culture environment parameters and external reference data, and performs standardization and structured processing;

[0044] Hybrid combination optimization module: The biased random bond genetic algorithm is used to encode and optimize the hybridization combination, and the optimal hybridization scheme is generated by combining the dual elite evolution mechanism and dynamic parameter adjustment strategy.

[0045] Cultivation Condition Prediction Module: Constructs a dynamic environmental model using a hybrid inverse reinforcement learning algorithm and searches for the optimal cultivation strategy that maximizes iron content in a virtual environment;

[0046] Strain performance evaluation module: Based on the analytic hierarchy process, a multi-index comprehensive scoring system is constructed to identify abnormal samples and trigger model backtracking and adjustment, and output high-performance strains to enter the next round of breeding.

[0047] Compared with the prior art, the present invention achieves the following beneficial effects:

[0048] 1. This invention proposes a random-bond chromosome and a dual-elite evolutionary mechanism, constructing a biased random-bond genetic algorithm: by encoding hybridization combinations as continuous real-number random bond vectors, a chromosome expression method decoupled from the problem structure is constructed. A dual-elite mechanism of elite retention + biased mating is introduced to improve the retention and genetic ability of high-quality solutions. Biased crossover and mutation operations are used to avoid premature convergence, achieving efficient global search of the strain combination space.

[0049] 2. This invention proposes an adaptive parameter dynamic adjustment strategy: as the evolutionary process progresses, the system dynamically adjusts parameters such as the elite ratio, mutation probability, and population size. This improves the strategy adaptability in the early exploration and later convergence stages, enhancing the algorithm's stability and generalization ability.

[0050] 3. This invention proposes a model-based hybrid inverse reinforcement learning algorithm: It comprehensively utilizes expert demonstration data and learner interaction data to construct a hybrid environment model. A neural network is used to fit the state transition probability distribution, and a closed-loop learning mechanism of "reward function → policy → model" is formed through policy optimization and reward function iteration. This efficiently simulates the adjustment behavior of culture medium formulation and environmental conditions in a virtual model environment, automatically searching for the optimal culture strategy.

[0051] 4. This invention proposes a regularized loss function design to prevent model overfitting: During dynamic model training, the loss function incorporates weight decay (L2 norm regularization) and gradient clipping mechanisms. This avoids overfitting the model under small sample or expert data distributions and improves its generalization ability to new samples and complex environments.

[0052] 5. This invention proposes a strain performance evaluation method based on a comprehensive scoring mechanism and penalty items: A multi-index evaluation system based on AHP is constructed, weighting and integrating iron content (50%), yield (30%), and growth period (20%). A standardized scoring mechanism and penalty items are introduced to automatically identify low-performance or substandard strains, and the upstream optimization process is traced back to iteratively improve breeding strategies.

[0053] 6. Based on the evaluation of a large number of black fungus germplasm resources, this invention selectively chooses strains with high iron content as hybridization parents, making it easier to create super-fertile high-iron strains. Combined with the addition of culture media with appropriate exogenous iron ion concentrations, high-iron-content strains are then cultured. This method is more targeted and efficient than using hybridization breeding or mutation breeding techniques alone, thus improving breeding efficiency. This invention not only improves the genetic traits of black fungus but also, by altering culture conditions, doubly enhances the high-iron trait, significantly improving the accuracy of strain selection and providing a new approach for the breeding of functional edible fungi strains.

[0054] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0055] The above and other features, advantages, and aspects of the various embodiments of the present invention will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the invention. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0056] Figure 1 This is a schematic diagram of a method for hybridization and selection optimization of high-iron strains of black fungus based on big data analysis according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of a module of a hybridization and optimization system for high-iron strains of black fungus based on big data analysis, according to an embodiment of the present invention.

[0058] Figure 3 The present invention describes the monokaryotic mycelium of the black fungus strain JAUA628 under an optical microscope.

[0059] Figure 4 The present invention relates to the binucleate mycelium of the black fungus strain JAUA628 under an optical microscope;

[0060] Figure 5 The black fungus strain JAUA628 of this invention is antagonistic to its parent strain;

[0061] Figure 6 The fungus strain JAUA628 of this invention is the fungus that produces fruiting fruit. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of 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.

[0063] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0064] This invention starts with strains of black fungus that have high iron content, and creates super-iron-loving strains through hybridization breeding. Then, it combines the culture medium with exogenous iron ion supplementation to cultivate high-iron black fungus strains, which is an effective way to efficiently obtain high-iron black fungus strains.

[0065] The high-iron black fungus strain JAUA628 was developed using the high-iron-content wild black fungus strain A356 (iron content: 161.4 mg / kg) and the high-yield cultivated strain A592 as parents. Spores were collected, and single-spore isolation was performed to obtain monokaryotic strains. Single-single hybridization was then conducted to create true hybrid strains. The hybrid strain AA48 with the highest iron content and superior to the parent strain was selected. The monokaryotic strain obtained from spore collection was backcrossed with the monokaryotic strain of parent A356, and the hybrid strain JAUA628 with high iron content and superior to the parent strain was selected, with an iron content of 194.3 mg / kg, exceeding the parent strain by 20%. When the hybrid strain JAUA628 was produced in a culture medium with a formula of 78% sawdust, 20% wheat bran, 1% lime, 1% gypsum, 2.0 g / kg ferrous sulfate, and 60% moisture content, the iron content of the hybrid strain JAUA628 was increased to 221 mg / kg.

[0066] The strain JAUA628 of *Auricularia auricula-judae* was deposited at the China Center for Type Culture Collection on April 19, 2024, with accession number CCTCCM2024717.

[0067] The high-speed iron-fed black fungus strain of this invention was created by combining genetic breeding with appropriate cultivation methods to improve the strain based on its unique traits. However, research on the breeding of functional varieties of black fungus has not yet been reported. Other functional edible fungi are mostly developed by changing culture conditions and utilizing the enrichment capacity of edible fungi, without altering their genetic traits, which has certain limitations.

[0068] Based on the evaluation of a large number of black fungus germplasm resources, this invention selectively chooses strains with high iron content as hybridization parents, making it easier to create super-fertile high-iron strains. Combined with the addition of culture media with appropriate exogenous iron ion concentrations, high-iron-content strains are then cultured. Compared with hybridization breeding or mutation breeding techniques alone, this method is more targeted and efficient, improving breeding efficiency.

[0069] This invention not only improves the genetic traits of black fungus but also enhances the high-iron trait by altering culture conditions, significantly improving the accuracy of strain selection. It provides a new approach for the breeding of functional strains of edible fungi.

[0070] This invention utilizes artificial intelligence to construct a hybridization and optimization system for high-iron strains of Auricularia auricula-judae based on big data analysis. It forms a closed-loop automated optimization process from data acquisition, combination optimization, environmental prediction to performance evaluation, providing an intelligent decision-making tool for the breeding of edible and medicinal fungi such as Auricularia auricula-judae. It significantly improves efficiency and breeding accuracy in hybridization combination optimization, environmental prediction modeling, and performance evaluation.

[0071] Figure 1 This diagram illustrates a flowchart of a method for hybridization and optimization of iron-rich strains of Auricularia auricula-judae based on big data analysis, according to an embodiment of the present invention. Figure 1 As shown, a hybridization and optimization method 100 for high-iron strains of Auricularia auricula-judae based on big data analysis includes:

[0072] Step S1: Collect basic data of the strain, hybridization process data, culture environment parameters and external reference data, and perform standardization and structuring processing;

[0073] Step S1 is used for multi-source data acquisition and preprocessing. To construct an intelligent hybridization and optimization system for high-iron strains of *Auricularia auricula-judae*, a complete, standardized, and structured data foundation is first required. Systematic acquisition and cleaning of multi-source heterogeneous data provides high-quality support for subsequent hybridization optimization, model prediction, and performance evaluation. The main steps include:

[0074] Step S1.1: Collection of four types of data

[0075] Step S1.1.1: Basic data collection of strains

[0076] Genetic background and agronomic traits of key parental strains are collected as basic inputs for hybridization design.

[0077] Genetic information, such as genotype data of parents A356 and A592, clamp connection characteristics of monokaryotic hyphae, and karyotype information, is used to help infer the genetic expression capacity of fusion offspring.

[0078] Iron content determination: For example, the iron content of wild black fungus strain A356 is 161.4 mg / kg, which is used as a reference baseline for offspring over-parental performance.

[0079] Agronomic trait data, including mycelial growth rate, yield performance, and growth cycle (such as fruiting time and maturity time), are used for later comprehensive performance evaluation.

[0080] Step S1.1.2: Data Acquisition During Hybridization

[0081] Data was collected for key steps in each round of hybridization experiments, including microscopic operation records and experimental result indicators.

[0082] Single-spore isolation data: Record the number of mononuclear strains isolated from the fruiting bodies in each round of experiments (e.g., 18 in the first round) to provide a diversity basis for subsequent combinatorial design.

[0083] Hybridization combination design: Record the number of combinations generated in each round of hybridization (e.g., 81 combinations are generated in the first round), and combine them with naming rules to make them easy to track and manage.

[0084] Antagonistic reaction experimental data: Images of antagonistic reactions and response levels (such as fusion, non-fusion, and excessive reaction) between different hyphal combinations were collected to eliminate non-true hybridization combinations or homologous parent groups.

[0085] Agronomic trait data, including mycelial growth rate, yield performance, and growth cycle (such as fruiting time and maturity time), are used for later comprehensive performance evaluation.

[0086] Step S1.1.3: Collection of culture environment and condition parameters

[0087] We collected key environmental factors that influence strain growth and iron enrichment to support subsequent reinforcement learning environment modeling and simulation.

[0088] Culture medium formulation composition: Record the basic materials (such as 78% hardwood sawdust, 20% wheat bran, lime / gypsum and other additives) as input parameters for the culture environment.

[0089] Iron ion concentration setting: This includes a set concentration gradient (e.g., 1.0-2.5 g / kg ferrous sulfate), which is used as an input variable to correlate the target iron content in the prediction model.

[0090] Temperature and humidity conditions: including continuous or phased environmental control data such as temperature during the mycelial growth stage (22-28℃) and humidity during the ear emergence stage (85-90%).

[0091] Agronomic trait data, including mycelial growth rate, yield performance, and growth cycle (such as fruiting time and maturity time), are used for later comprehensive performance evaluation.

[0092] Step S1.1.4: External Reference Data Acquisition

[0093] Introduce the technical routes and successful cases of other edible and medicinal fungi (such as Ganoderma lucidum and Cordyceps militaris) in functional breeding as prior information for system design and model training.

[0094] Functional breeding methods include: breeding processes for polysaccharide enrichment and selenium adsorption capacity, and strategies for enhancing metabolic pathways.

[0095] Breeding evaluation index system: Refer to the multi-objective evaluation method used in its functional evaluation, compare and absorb its effective experience.

[0096] Agronomic trait data, including mycelial growth rate, yield performance, and growth cycle (such as fruiting time and maturity time), are used for later comprehensive performance evaluation.

[0097] Step S1.2: Data Preprocessing Flow

[0098] To ensure data availability and model training effectiveness, the collected multi-source data undergoes standardization and structuring processing, including:

[0099] Missing value filling and outlier removal (e.g., iron content below 10 mg / kg is directly removed).

[0100] Numerical normalization processing (applicable to various continuous indicators, such as temperature, concentration, etc.).

[0101] Categorical feature encoding (such as antagonistic reaction level, strain identifier).

[0102] Time series alignment and stage labeling (for reproductive period and dynamic simulation modeling).

[0103] Through the above-mentioned multi-dimensional data acquisition and preprocessing operations, the system has constructed a high-quality data foundation for strain breeding that integrates genetics, physiology, environment, and process behavior, laying a solid digital platform foundation for subsequent algorithm optimization and performance improvement.

[0104] Step S2: Use a biased random bond genetic algorithm to encode and optimize the hybridization combination, and combine a dual-elite evolutionary mechanism and a dynamic parameter adjustment strategy to generate the optimal hybridization scheme;

[0105] To effectively improve the efficiency of hybridization breeding of iron-rich strains of Auricularia auricula-judae and reduce ineffective combinations and experimental costs, this system, based on data acquisition, further introduces a biased random bond genetic algorithm to construct an intelligent hybridization optimization mechanism and process simulation model. By comprehensively considering two objectives—"iron content over-parent rate" and "agronomic trait stability"—the hybridization combination strategy and backcross generation number are dynamically optimized, significantly improving the screening accuracy of real hybrid offspring. Step S2 is used to realize hybridization combination optimization and process simulation, and the implementation of the biased random bond genetic algorithm includes:

[0106] The hybridization combination is encoded as a random key chromosome vector X represented by n-dimensional continuous real numbers;

[0107] The random bond chromosome vector X is decoded into a specific combination strategy and fitness score by an encoder. The objective function is constructed with iron content over-parent rate and agronomic trait stability as the evaluation core.

[0108] By introducing parental gene expression data, simulating gene recombination behavior during the fusion of different monokaryotic strains, predicting the probability distribution of iron content in offspring, and forming a dynamic hybridization simulation model;

[0109] Employing a dual-elite evolutionary mechanism, each generation retains the previous generation's fitness. There are elite individuals, and a biased crossover operation is performed with probability p>0.5: the offspring genes inherit the elite parent 1 gene with probability p, and inherit the non-elite parent 2 gene with probability 1-p.

[0110] As the number of iterations progresses, parameters such as the elite ratio and mutation probability are dynamically adjusted to adapt to the search progress.

[0111] Step S2 specifically includes the following steps:

[0112] Step S2.1: Hybrid Combination Optimization Strategy

[0113] Traditional hybridization methods involve a large number of combinations (e.g., 81 combinations in the first round), but the actual hybridization success rate is low (e.g., only 36 true hybrid strains were screened). Therefore, this system adopts the following strategy:

[0114] Antagonistic reaction screening: Based on the results of hyphal antagonistic reaction experiments, combinations with significant phylogenetic differences are preferentially retained, while homogeneous or non-fusion combinations are initially excluded.

[0115] Dynamic adjustment of backcross strategy: Based on the mid-term test results, the number of backcrosses of offspring with obvious iron-rich traits, such as "hybrid strain AA48 and wild black fungus strain A356-4 with high iron content", will be dynamically increased to ensure the genetic stability of the target trait.

[0116] Multi-objective optimization: By constructing a joint objective function, the improvement of iron content is balanced with the stable performance of agronomic traits such as yield and growth period.

[0117] Step S2.2: Construct a biased random key genetic algorithm

[0118] The system employs the Biased Random Key Genetic Algorithm (BRKGA) for combinatorial optimization. Its core feature is the combination of random key chromosomes and a dual-elite evolutionary mechanism, which has good search space coverage and global optimization efficiency.

[0119] The random-key chromosome vector is a virtual encoding designed in the biased random-key genetic algorithm to represent the optimization strategy of crossover combinations. It is a mathematical abstract model, with each dimension... For each breeding decision variable, a decoder can transform it into an actionable hybridization breeding strategy (e.g., parental selection, number of backcrosses, recommended environmental parameters such as iron concentration in the culture medium), ultimately outputting the optimal strategy to guide experimental design. The random bond chromosome vector is defined as a real-number vector of dimension n representing each individual.

[0120]

[0121] X: Random key chromosome vector, used to encode the selection order and strategy parameters of strain combinations; n is the encoding dimension, determined by the number of decision variables to be optimized (e.g., parental selection, number of backcrosses, environmental parameter weights, etc.). The i-th random key represents the priority or proportion allocation parameter in the decoder. It is a continuous real number in the interval [0,1], representing the probability weight or priority of a certain decision. The decoder decodes the random key vector X into a specific combination of policy and fitness score, for example, by threshold decoding (e.g., when...). The strategy for generating hybridization combinations is to select the i-th parent. The objective function is constructed with iron content and trait stability as the core evaluation criteria:

[0122]

[0123] : Fitness score corresponding to random key chromosome vector X; Iron content exceeding parental rate: the predicted rate at which offspring have iron content higher than the average of their parents; : Stability of agronomic traits, a trait consistency index fitted based on data such as ear emergence date and yield; α, β: Weight parameters of the objective function, set according to task requirements. Step S2.3: Dynamic simulation of the hybridization process

[0124] By incorporating parental gene expression data, simulating gene recombination behavior during the fusion of different monokaryotic strains, and predicting the probability distribution of iron content in offspring, a dynamic hybridization simulation model is formed.

[0125]

[0126] : Expected iron content of offspring in the k-th combination; : Expression level of the j-th iron-related gene; The weight factor of the j-th gene in the k-th combination is obtained by fitting a linear regression model of the parental gene expression data, and the model training uses the least squares method. : The number of target genes involved in the simulation; The summation symbol for subscripts j from 1 to g indicates the summation of the weighted expression values ​​of all iron-related genes; : Mathematical expectation operation, representing the weighted average of all possible offspring gene combinations under the condition of uncertainty in gene expression and recombination.

[0127] Step S2.4: Evolutionary Process and Dynamic Parameter Adjustment

[0128] During the algorithm iteration process, a dual-elite mechanism and an adaptive parameter adjustment strategy are adopted:

[0129] Elite replication: Preserving fitness before each generation individual

[0130] Mutation operation: generating Randomized chromosomes to increase diversity

[0131] Bias crossover: based on probability Bias selection of elite genes generates the next generation of individuals:

[0132]

[0133] The i-th gene value of the offspring is the coding value at the i-th position in the new individual generated after mating, and it is a component of the new generation chromosome. : Gene from elite parent 1, indicating that the gene locus was selected from individuals with high fitness (i.e., good performance) in the current population; : Gene from non-elite parent 2, representing the gene locus selected from non-optimal individuals in the current population; p: probability of retaining elite genes, indicating the probability that the i-th gene will be directly inherited from an elite individual when generating a new generation of chromosomes. Since p > 0.5, this strategy has a clear "elite bias" property, strengthening the transmission of superior genes; 1 - p: probability of non-elite genes being selected, used to preserve population diversity and avoid getting trapped in local optima. Optionally, the bias probability p increases linearly from 0.6 to 0.8.

[0134] With the number of iterations As the search progresses, the system dynamically adjusts parameters such as the elite ratio and mutation probability to adapt to the search progress, as shown in the following formula:

[0135]

[0136] : The number of elite individuals at the t-th algorithm iteration; Maximum number of iterations; , The set minimum and maximum elite ratio thresholds. initial value It is 0.2. The range of dynamic adjustment is: , .

[0137] Step S2.5: Jitter and Reset Mechanism

[0138] To avoid the algorithm getting trapped in local optima, by introducing... Proportionate mutant individuals (completely randomly generated new solutions) enhance population diversity. Elite individuals are slightly perturbed using a jitter operator, and are probabilistically... Apply a perturbation, expressed as:

[0139]

[0140] Disturbance amplitude parameters; Uniformly distributed random variables; : Perturbation probability; : The i-th random key value, with the same meaning as in step S2.2.

[0141] When there is no fitness improvement for several consecutive generations, a reset operation is performed to remove non-elite individuals and regenerate new random individuals (random chromosomes) to avoid premature convergence.

[0142] Step S3: Construct a dynamic environmental model using a hybrid inverse reinforcement learning algorithm with a model, and search for the optimal cultivation strategy that maximizes iron content in the virtual environment;

[0143] Step S3 analyzes the nonlinear relationship between the iron ion concentration (range: 0–2.5 g / kg) in the culture medium and the iron content of the fruiting bodies (range: 194.3–221 mg / kg), constructs an iron content prediction model, and automatically searches for the optimal culture formula based on a model-based hybrid inverse reinforcement learning algorithm, with "maximizing the iron content of the fruiting bodies" as the reward objective. The specific process of intelligent prediction of culture conditions in step S3 includes:

[0144] Step S3.1: Construct a hybrid inverse reinforcement learning model

[0145] Step S3.1 constructs a dynamic environmental model using a hybrid inverse reinforcement learning algorithm to search for the optimal cultivation strategy that maximizes iron content in a virtual environment. This includes: constructing a hybrid dataset using expert demonstration data and online learner data; and fitting the environmental state transition probabilities using a neural network to obtain the dynamic environmental model. Furthermore, a closed-loop learning mechanism is formed by strategy optimization and reward function iteration to efficiently simulate the adjustment behavior of culture medium formulation and environmental conditions in a virtual model environment. Specifically:

[0146] The input data is for expert demonstration purposes. (High-quality priors) and learner online data (Real-time correction), mixed to form the training set ( (For union operations). Through data mixing, expert demonstrations are utilized. It provides high-quality state-action pairs, reduces the policy search space, and avoids the inefficiency of random exploration; at the same time, it utilizes learner trajectories. Correcting model errors enhances adaptability to environmental dynamics, improves the generalization ability of environmental dynamic predictions, and prevents model overfitting to expert data. Expert demonstration data. This includes: artificially optimized culture medium formula (e.g., ferrous sulfate 1.5–2.5 g / kg), temperature and humidity control curve (mycelium growth period 25±2℃), and harvesting time (early stage of spore release).

[0147] Neural networks (such as Transformer or multilayer perceptron) are used to simulate environmental dynamics. The state space is defined as S = {culture medium formula C, temperature and humidity T, iron ion concentration Fe, mycelial growth rate G, metabolite concentration M}, and the action space is defined as A = {adjusting Fe concentration ΔFe, changing temperature ΔT, modifying formula ratio ΔC}. A reward function is also used. ,in, The actual iron concentration of the fruiting body (e.g., JAUA628 has an iron content of 221 mg / kg at 2.0 g / kg culture medium); ΔT is the temperature difference before and after the action is performed (e.g., from 25°C to 28°C). ); ΔC represents the adjustment range of the culture medium composition (e.g., wheat bran content changes from 18% to 20%). ); For the corresponding weighting coefficients, The typical ratio is 20:1:0.5, ensuring that iron content contributes more than 95% of the total reward value, and the weight can be dynamically adjusted with each iteration stage, initially focusing on exploration. Lower), later stages focus on convergence ( Approaching 1.0). Predicting in state. Next action The next state distribution is then determined. The training objective is to minimize the KL divergence between the predicted and true state distributions, and the loss function is... Overfitting can be avoided by adding weight decay and gradient clipping:

[0148]

[0149] Model loss function, used to evaluate the current learning model. With real-world dynamics The gap; Mixed datasets, presented by experts Data generated with learners The combined dataset; : The learned dynamic model of the environment; s: the current state; a: the action; s': the next state; In a real-world environment, given a state and actions The probability distribution of the next state; The next state distribution predicted by the learning model is the neural network's estimation of environmental dynamics. These are model parameters; The regularization coefficient is used. Norm square, also known as regularization term, is used to constrain the parameter size. KL divergence measures the difference between two probability distributions.

[0150] This loss function can prevent the model from relying too heavily on the distribution of expert data and enhance its adaptability to complex environments with small samples.

[0151] Step S3.2: Intra-model policy search

[0152] Step S3.2 involves randomly selecting states from the expert demonstration dataset as the initial states for policy optimization. This is relevant in the environmental dynamics model. Within this framework, a gradient-free optimization algorithm or model predictive control method is used to optimize policy π to maximize the cumulative reward, and the reward function is updated through gradient ascent to generate a new policy π. This data is further fed back into the real environment, generating new data. This forms a closed data loop. Specifically:

[0153] Randomly sample states from expert trajectories This serves as the initial state for policy optimization within the model, simulating the starting point of expert behavior. In learning the model... In the context of the current reward function Optimization strategy Maximize cumulative rewards:

[0154]

[0155] Where π represents the policy, defining the agent's actions in a given state. Select action The probability distribution; The new strategy obtained in the (t+1)th iteration; The trajectory ξ is of the form of The state-action pair sequence is constructed through a policy. In environmental dynamics model The simulation sampling is generated; ξ represents a complete trajectory path, containing all steps from the initial state to the final state; The learned dynamic model of the environment is used to simulate the state transition process (i.e., predict the next state based on the current state and actions). In the t-th iteration, the reward function is applied to the state at the h-th step of the trajectory. and actions The score is used to measure the quality of the decision made in that step; H: trajectory length, which is the maximum number of steps or rounds in the simulation path, indicating that a trajectory contains a maximum of H state-action pairs (from...). arrive ; In the model Below, the trajectory obtained by sampling strategy π. Calculate the expected value, used to calculate the long-term average effect of statistics such as cumulative rewards; The optimization objective is to select the strategy that maximizes the expected cumulative reward from all possible strategies π, i.e., to find the optimal strategy.

[0156] Gradient-free optimization algorithms (such as cross-entropy methods and proximal policy optimization) or model predictive control methods can be used to efficiently iterate policies within the model, reducing frequent reliance on the real environment. Low-cost policy iteration in a virtual environment reduces the number of real experiments and utilizes reward functions. Dynamic guidance strategies approximate expert behavior.

[0157] Step S3.3: Update the reward function

[0158] We optimize the reward function through gradient ascent to achieve regret-free learning, gradually bringing it closer to an expert policy:

[0159]

[0160] in, The current reward function, used in the t-th iteration; The reward function updated after the (t+1)th iteration; The learning rate is used to control the update step size of the reward function. : Calculate the gradient of the reward function f; In the reward function Below, expert strategy Expected cumulative reward; In the reward function Below, the current strategy Expected cumulative reward; Expert strategies typically stem from human experience or high-quality demonstrations. : The current strategy, i.e. the strategy optimized in the t-th iteration.

[0161] Optionally, based on preliminary experimental verification, when the cumulative reward of strategy π... If the improvement rate is less than 1% after 5 consecutive iterations, and the contribution of further optimization to the improvement of iron content is less than 0.5%, then the optimization is terminated.

[0162] By continuously refining the reward function, learners' strategies can increasingly approximate expert behavior, achieving a closed-loop reinforcement learning process of "reward function update → strategy optimization → model correction".

[0163] The optimized strategy was applied to a real-world training system, new trajectory data was collected, and the new data was filtered for effectiveness, retaining high-reward and highly diverse trajectories for further training of the model and reward function. The final output results obtained through experiments are as follows:

[0164] Recommended optimal culture formula: 78% hardwood sawdust + 20% wheat bran + 1% lime + 1% gypsum + 2.0 g / kg ferrous sulfate (moisture content 60%), with a predicted iron content of 221 mg / kg.

[0165] Recommended environmental parameters: early stage of mycelial growth temperature 25-28℃, middle and late stage 22-24℃, ear emergence humidity 85-90%.

[0166] Step S3 utilizes inverse reinforcement learning to nest and iterate through environmental modeling (step 3.1), policy optimization (step 3.2), and reward updating (step 3.3), achieving low-cost trial and error in a virtual environment (reducing real breeding experiments) and self-correction of the reward function (avoiding the subjectivity of manually designed rewards). Furthermore, the deep integration of inverse reinforcement learning and the dynamic model forms a self-iterable decision-making closed loop, enabling intelligent prediction and autonomous optimization of black fungus cultivation conditions.

[0167] Step S4: Construct a multi-index comprehensive scoring system based on the analytic hierarchy process (AHP) to identify abnormal samples and trigger model backtracking adjustments, and output high-performance strains to enter the next round of breeding.

[0168] Step S4 comprehensively evaluates the performance of candidate strains in terms of iron content, yield, and stress resistance (such as growth period), screens high-performance strains, and backtracks to adjust the model strategy when standards are not met, thereby improving breeding efficiency. Step S4, the process of strain performance evaluation and iteration, specifically includes:

[0169] Step S4.1: Construction of a multi-indicator comprehensive evaluation system

[0170] The Analytic Hierarchy Process (AHP) was used to assign weights to each indicator: iron content 50%, yield 30%, and growth period (representing stress resistance) 20%. The comprehensive score for each strain was calculated as follows:

[0171] Overall score = 0.5 × iron content score + 0.3 × yield score + 0.2 × reproductive period score

[0172] The scores for each item are standardized. Iron content score / yield score = (measured value - minimum value) / (maximum value - minimum value) × 100. For example, in the iron content score, the minimum value = 100 mg / kg and the maximum value = 250 mg / kg.

[0173] Reproductive period score = 1 - |(Reproductive period length of the strain to be evaluated - preset optimal reproductive period length) / preset optimal reproductive period length|.

[0174] Step S4.2: Automatic identification and removal of abnormal samples

[0175] The following types of samples will be identified as abnormal and automatically removed:

[0176] (1) The iron content is lower than the average level of the parents.

[0177] (2) The iron content is lower than the target threshold of 194.3 mg / kg

[0178] (3) Abnormalities during the reproductive period (e.g., deviation from the normal range ±15%)

[0179] (4) Missing indicators or obvious data errors

[0180] The exclusion logic is as follows: if the iron content is <194.3mg / kg or lower than the average value of the parent strain, the sample of this strain will be excluded.

[0181] Step S4.3: Iterative optimization and model backtracking adjustment

[0182] If most strains fail to meet the standards in a certain round of evaluation (e.g., the overall score is below 0.6 or the iron content is insufficient), the system will automatically go back to the previous step (e.g., parental selection or hybridization design) and perform the following optimizations:

[0183] (1) Increase the weight of high-speed rail markers to enhance the selection bias for high-speed rail traits;

[0184] (2) Re-screen the parents and remove materials with excessively low basic iron content;

[0185] (3) Update the parameters of the genetic algorithm or recommendation model to optimize the prediction quality of mating combinations.

[0186] The parameter adjustment logic can be expressed by adding a penalty term (which only takes effect when the iron content does not meet the standard). That is, if the strain's compliance rate is lower than the threshold, the weight of the high iron trait is increased or a penalty term is added. The total loss is calculated as follows:

[0187] Total loss = Original loss + Penalty coefficient × (Average iron content of parents - Actual iron content) 2

[0188] The penalty coefficient is dynamically adjusted based on the degree to which the iron content fails to meet the standard.

[0189] According to step S4.3, when the overall score of the strain is lower than the preset threshold or the iron content is not up to standard, the system automatically backtracks to the parent selection or hybridization design stage and updates the parameters of the genetic algorithm or recommendation model to optimize the prediction quality of mating combinations.

[0190] Step S4.4: Output and Next Round of Evaluation

[0191] The top 10% of strains by overall score are retained as excellent candidates to advance to the next round of experiments; other samples are used for model training or elimination. The system records and outputs the following information: a score distribution chart for each round; a list of eliminated samples and the reasons for elimination; and a record of adjustments to model parameters and weights to facilitate tracking the optimization path.

[0192] The above is an introduction to the method embodiments. The following system embodiments will further illustrate the solution of the present invention.

[0193] Figure 2 This is a schematic diagram of a module of a hybridization and optimization system for high-iron strains of black fungus based on big data analysis, according to an embodiment of the present invention. Figure 2 As shown, the present invention provides a hybridization and optimization system 200 for high-iron strains of black fungus based on big data analysis, comprising:

[0194] Data acquisition and preprocessing module 210: Collects basic data of strains, hybridization process data, culture environment parameters and external reference data, and performs standardization and structuring processing;

[0195] Hybrid combination optimization module 220: The biased random key genetic algorithm is used to encode and optimize the hybrid combination, and the optimal hybridization scheme is generated by combining the dual elite evolution mechanism and dynamic parameter adjustment strategy.

[0196] Cultivation Condition Prediction Module 230: Constructs a dynamic environmental model using a model-based hybrid inverse reinforcement learning algorithm to search for the optimal cultivation strategy that maximizes iron content in a virtual environment;

[0197] Strain performance evaluation module 240: Based on the analytic hierarchy process, a multi-index comprehensive scoring system is constructed to identify abnormal samples and trigger model backtracking and adjustment, and output high-performance strains to enter the next round of breeding.

[0198] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0199] Furthermore, based on the above-mentioned method for hybridization and selection optimization of high-iron strains of Auricularia auricula-judae based on big data analysis and the optimal culture strategy determined by the system, the present invention conducted breeding experiments, and the experimental results are as follows:

[0200] 1. First round of high-iron hybrid strain creation: Using the high-iron-content wild black fungus strain A356 (iron content 161 mg / kg) and the high-yield cultivated strain A592 as parents, spores were collected after fruiting, and single-spore isolation was performed to obtain monokaryotic strains. The authenticity of the monokaryotic strains was identified by observing the presence or absence of clamp connections in the mycelium under an optical microscope. Figure 3 and Figure 4 Eighteen monokaryotic strains were prepared from the two parents, and single-cross hybridization was performed to prepare 81 hybrid combinations. These combinations were then subjected to antagonistic reactions with the two parents. Figure 5 and Figure 6 Thirty-six real hybrid strains were screened out and subjected to fruiting tests. The iron content of the fruiting bodies was measured and is shown in Table 1. The strain with the highest iron content was AA48, with a content of 173 mg / kg.

[0201] Table 1 Iron content of the first round of hybrid strains

[0202]

[0203] 2. Creation of the second round of high-iron hybrid strains: Using the screened high-iron hybrid strain AA48 as the parent, spores were collected to obtain monokaryotic strains. After the authenticity of the monokaryotic strains was identified, a total of 53 monokaryotic strains were prepared. Backcrossing experiments were conducted with the parent A356-4 of hybrid strain AA48 to obtain 24 hybrid strains. After antagonistic reactions with the two parents, 17 authentic hybrid strains were screened out. Fruiting experiments were conducted to detect the iron content and agronomic traits of the fruiting bodies (Tables 2 and 3). Finally, hybrid strain JAUA628 was selected as having the highest iron content, at 194.3 mg / kg, exceeding the parent by 20%.

[0204] Table 2 Iron content of the second round of hybrid strains

[0205]

[0206] Table 3 Agronomic traits of hybrid strains

[0207]

[0208] 3. Screening of culture medium formula:

[0209] 1) Experimental Treatment: Iron ion concentration gradients of 1.0 g / kg, 1.5 g / kg, 2.0 g / kg, and 2.5 g / kg were set in the culture medium. Different concentrations of ferrous sulfate solution were added to the traditional culture medium formula (78% hardwood sawdust, 20% wheat bran, 1% lime, and 1% gypsum). The specific iron ion concentration gradients are shown in Table 4. The corresponding concentrations of ferrous sulfate were weighed and dissolved in the mixing water. After complete dissolution, the solution was mixed into the culture medium, thoroughly mixed, and then bagged.

[0210] Table 4. Details of Iron Ion Addition at Different Concentrations

[0211]

[0212] 2) Preparation and cultivation of mushroom bags: Use 17×33cm polyethylene plastic bags, autoclave for 2 hours, and inoculate after the bags have cooled to room temperature; the inoculation amount is 10-15g / bag. For mycelial growth cultivation, the temperature is 25-28℃ in the early stage and 22-24℃ in the middle and late stages. After the bags are fully filled, allow them to mature for 10 days.

[0213] 3) Ear emergence management: Drill holes to promote germination, temperature 15-20℃, relative humidity 70-80%. Once primordia appear at the drilled holes, manage ear emergence by maintaining a greenhouse temperature of 15-24℃ and relative humidity of 85-90%. Spray water 2-3 times a day to maintain humidity and provide diffused light. When the ear matures, the outer edge becomes thinner and a small number of spores are ejected, indicating that the ear is mature and ready for harvest.

[0214] 4) Determination of iron content in the fruiting bodies of strain JAUA628

[0215] Iron content was determined in the fruiting bodies of strain JAUA628 in different treatment groups and control groups using inductively coupled plasma atomic emission spectrometry.

[0216] 5) Screening of iron-rich culture medium formula for strain JAUA628

[0217] Based on the iron content of fruiting bodies in different treatment groups and the control group, the optimal iron ion concentration in the culture medium was selected. The results showed that the iron content of fruiting bodies of strain JAUA628 increased with increasing iron ion concentration in the culture medium, reaching a maximum of 221 mg / kg at a concentration of 2.0 g / kg (Table 5). Therefore, the optimal iron-rich culture medium formula for strain JAUA628 is: 78% hardwood sawdust, 20% wheat bran, 1% lime, 1% gypsum, 60% moisture content, and 2.0 g / kg ferrous sulfate (60% moisture content).

[0218] Table 5 Iron content of fruiting bodies of strain JAUA628

[0219]

[0220] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0221] It should also be noted that, in the embodiments of this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0222] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined in the embodiments of this application may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown in this application, but is to be accorded the widest scope consistent with the principles and novel features disclosed in the embodiments of this application.

Claims

1. A method for hybridization and optimization of iron-rich strains of Auricularia auricula-judae based on big data analysis, characterized in that, include: Step S1: Collect basic data of the strain, hybridization process data, culture environment parameters and external reference data, and perform standardization and structuring processing; Step S2: Use a biased random bond genetic algorithm to encode and optimize the hybridization combination, and combine a dual-elite evolutionary mechanism and a dynamic parameter adjustment strategy to generate the optimal hybridization scheme; Step S3: Construct a dynamic environmental model using a hybrid inverse reinforcement learning algorithm with a model, and search for the optimal cultivation strategy that maximizes iron content in the virtual environment; Step S4: Construct a multi-index comprehensive scoring system based on the analytic hierarchy process (AHP) to identify abnormal samples and trigger model backtracking adjustments, outputting high-performance strains for the next round of breeding, including: The Analytic Hierarchy Process (AHP) was used to assign weights to each indicator: iron content 50%, yield 30%, and growth period 20%. The comprehensive score for each strain was calculated as follows: Overall score = 0.5 × iron content score + 0.3 × yield score + 0.2 × reproductive period score Each score is standardized. When the overall score of the strain is lower than the preset threshold or the iron content is not up to standard, the system automatically backtracks to the parent selection or hybridization design stage, updates the parameters of the genetic algorithm or recommendation model, and optimizes the prediction quality of mating combinations. If the strain compliance rate is lower than the threshold, the weight of the high-iron trait is increased or a penalty term is added. The total loss is calculated as follows: Total loss = Original loss + Penalty coefficient × (Average iron content of parents - Actual iron content) 2 The penalty coefficient is dynamically adjusted based on the degree to which the iron content fails to meet the standard.

2. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 1, characterized in that, The implementation of the biased random key genetic algorithm in step S2 includes: The hybridization combination is encoded as a random key chromosome vector X represented by n-dimensional continuous real numbers; The random bond chromosome vector X is decoded into a specific combination strategy and fitness score by an encoder. The objective function is constructed with iron content over-parent rate and agronomic trait stability as the evaluation core. By introducing parental gene expression data, simulating gene recombination behavior during the fusion of different monokaryotic strains, predicting the probability distribution of iron content in offspring, and forming a dynamic hybridization simulation model; Employing a dual-elite evolutionary mechanism, each generation retains the previous generation's fitness. There are elite individuals, and a biased crossover operation is performed with probability p>0.5: the offspring genes inherit the elite parent 1 gene with probability p, and inherit the non-elite parent 2 gene with probability 1-p. As the number of iterations progresses, parameters including the elite ratio and mutation probability are dynamically adjusted to adapt to the search progress.

3. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 2, characterized in that, in, The biased random key genetic algorithm also includes: Jittering mechanism: applying probability to elite individuals Apply a disturbance; Reset mechanism: When there is no improvement in fitness after a series of preset generations, remove non-elite individuals and regenerate random chromosomes.

4. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 3, characterized in that, The objective function is: ； in, : Fitness score corresponding to chromosome X; Iron content exceeding parental rate, which is the percentage of offspring predicted to have iron content higher than the average of the parents; : Stability index of agronomic traits; α, β: Weight parameters of the objective function, set according to task requirements.

5. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 4, characterized in that, Step S3: Constructing a dynamic environmental model using a hybrid inverse reinforcement learning algorithm, and searching for the optimal cultivation strategy to maximize iron content in a virtual environment, including: A hybrid dataset was constructed using expert demonstration data and online learner data. A neural network is used to fit the environmental state transition probabilities to obtain an environmental dynamic model. Furthermore, a closed-loop learning mechanism is formed by strategy optimization and reward function iteration to efficiently simulate the adjustment behavior of culture medium formulation and environmental conditions in a virtual model environment.

6. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 5, characterized in that, Step S3 further includes: Randomly select states from the expert demonstration dataset as the initial states for policy optimization; In environmental dynamics model Within this framework, a gradient-free optimization algorithm or model predictive control method is used to optimize policy π to maximize the cumulative reward, and the reward function is updated through gradient ascent to generate a new policy. This data is further fed back into the real environment, generating new data. This forms a closed data loop.

7. The method for hybridization and optimization of high-iron strains of Auricularia auricula-judae based on big data analysis according to claim 6, characterized in that, Iterative formula for reward function: ； The current reward function, used in the t-th iteration; The reward function updated after the (t+1)th iteration; The learning rate is used to control the update step size of the reward function. : Calculate the gradient of the reward function f; In the reward function Below, expert strategy Expected cumulative reward; In the reward function Below, the current strategy Expected cumulative reward; Expert strategies, derived from human experience or high-quality demonstrations; : The current strategy, i.e. the strategy optimized in the t-th iteration.

8. A hybridization and optimization system for high-iron strains of Auricularia auricula-judae based on big data analysis, characterized in that, include: Data acquisition and preprocessing module: Collects basic data of strains, hybridization process data, culture environment parameters and external reference data, and performs standardization and structured processing; Hybrid combination optimization module: The biased random bond genetic algorithm is used to encode and optimize the hybridization combination, and the optimal hybridization scheme is generated by combining the dual elite evolution mechanism and dynamic parameter adjustment strategy. Cultivation Condition Prediction Module: Constructs a dynamic environmental model using a hybrid inverse reinforcement learning algorithm and searches for the optimal cultivation strategy that maximizes iron content in a virtual environment; The strain performance evaluation module: Based on the analytic hierarchy process (AHP), a multi-index comprehensive scoring system is constructed to identify abnormal samples and trigger model backtracking adjustments. High-performance strains are then output for the next round of breeding, including: The Analytic Hierarchy Process (AHP) was used to assign weights to each indicator: iron content 50%, yield 30%, and growth period 20%. The comprehensive score for each strain was calculated as follows: Overall score = 0.5 × iron content score + 0.3 × yield score + 0.2 × reproductive period score Each score is standardized. When the overall score of the strain is lower than the preset threshold or the iron content is not up to standard, the system automatically backtracks to the parent selection or hybridization design stage, updates the parameters of the genetic algorithm or recommendation model, and optimizes the prediction quality of mating combinations. If the strain compliance rate is lower than the threshold, the weight of the high-iron trait is increased or a penalty term is added. The total loss is calculated as follows: Total loss = Original loss + Penalty coefficient × (Average iron content of parents - Actual iron content) 2 The penalty coefficient is dynamically adjusted based on the degree to which the iron content fails to meet the standard.

Images