Method and system for regulating the microflora of the soil

Abstract

The application discloses a soil micro-ecological flora control method and system, and belongs to the technical field of microorganism flora control, and comprises the following steps: continuously collecting real-time environmental data based on a target soil area, combining future climate prediction data, generating a predicted scenario sequence by using Monte Carlo simulation; constructing a microorganism community composition table, screening core strains, and performing macro-genome sequencing; taking the core strains as game participants, constructing a participant strategy set, calculating a revenue matrix library, setting an environmental correction coefficient for each scenario, performing scenario-based game solving, and generating a balance summary table; constructing a control response prediction model, setting multiple optimization objectives, introducing robustness constraints and expert rules, generating a candidate control scheme, and performing optimal scheme solving, which can not only ensure that the control measures can achieve the expected effect by guiding interspecific competition, but also avoid destroying the original flora balance.

Description

Technical Field

[0001] This invention relates to a method and system for regulating soil microbial communities, belonging to the field of microbial community regulation technology. Background Technology

[0002] With the advancement of modern intensive agricultural production models, unreasonable agricultural operations have led to continuous damage to the soil microecological balance, disrupting the original structure of the soil microbial community, and subsequently undermining the stable structure of the soil microbial community. Pathogens accumulate in the soil, causing soil acidification and compaction, which in turn leads to frequent occurrences of soil-borne diseases, continuous degradation of soil fertility, and significant inhibition of crop growth.

[0003] A Chinese patent application with publication number CN119005423A discloses a method and system for regulating soil microbial communities in vineyards, comprising: obtaining soil samples from the vineyard based on the grape planting area; analyzing the soil samples to obtain soil data; preprocessing the soil data to obtain a preprocessed dataset; constructing a fully connected neural network; improving the fully connected neural network and training it with a sample set to obtain a microbial community prediction model; inputting the preprocessed dataset into the microbial community prediction model for calculation to obtain microbial community prediction results; and regulating the soil microbial communities in the vineyards based on the microbial community prediction results.

[0004] Although existing technologies can accurately predict the distribution and potential changes of microbial communities from soil data, they do not consider the interspecific competition mechanisms of soil microorganisms and ignore the interactions between different species, such as resource competition and antagonistic inhibition. This leads to deviations between the model predictions and the actual evolution of microbial communities in the soil. Consequently, regulatory measures based on the predictions not only fail to achieve the expected results but also disrupt the original balance of the soil microbial community, damage the micro-ecological environment, and weaken the effectiveness and scientific validity of the regulatory schemes. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for regulating soil microbial communities, quantifying the functions of core microbial species and environmental scenarios, using game theory for scenario-based competition, embedding regulatory response prediction, and employing a multi-objective robust optimization scheme to solve the problems of prediction bias and regulatory failure caused by neglecting interspecific interactions.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Methods for regulating soil microbial communities include:

[0008] Based on the target soil region, real-time environmental data is continuously collected and combined with future climate prediction data to generate a predicted scenario sequence using Monte Carlo simulation.

[0009] A table of microbial community composition was constructed, and core bacterial species were screened and metagenomic sequencing was performed.

[0010] Using the core bacterial species as game participants, a set of participant strategies is constructed, a payoff matrix library is calculated, and an environmental correction coefficient is set for each scenario. The game is then solved in different scenarios to generate an equilibrium summary table.

[0011] A regulation response prediction model is constructed, multiple optimization objectives are set, robust constraints and expert rules are introduced, candidate regulation schemes are generated, and the optimal scheme is solved.

[0012] Specifically, the steps for screening core bacterial strains include:

[0013] Set a secondary division threshold to divide the target soil area into sampling zones and construct a sampling point distribution map;

[0014] Soil sample sets are generated by collecting samples based on environmental sampling depth.

[0015] DNA was extracted from soil samples and combined with high-throughput sequencing to obtain the raw sequence data of the soil samples. The data was then filtered according to quality control standards to generate a table of microbial community composition.

[0016] Calculate the relative abundance of each species, sort them in descending order, and generate a species abundance ranking table.

[0017] The relative abundances after sorting are summed to screen for core strains and form a list of core strains.

[0018] Specifically, the steps for screening core strains also include:

[0019] The relative abundance of each core microbial species in a single soil sample was extracted, and invalid core microbial species in a single soil sample were screened out by an effective abundance threshold.

[0020] The number of invalid core microbial species in a single soil sample was counted, and the coverage rate of core microbial species was calculated.

[0021] Based on the effective coverage threshold, qualified samples are screened to generate partitioned sequencing samples;

[0022] Based on functional database comparison, a functional trait table of core strains is generated, including resource uptake data, antagonistic inhibition data, and nutrient transformation data.

[0023] Specifically, the steps for calculating the revenue matrix library include:

[0024] Using the core bacterial species as game participants, assign participant identifiers to form a basic game table;

[0025] Set resource weight coefficients, average and sum the resource intake data, and calculate the resource intake index;

[0026] Calculate the antagonistic weighting coefficient, average and sum the antagonistic inhibition data, and calculate the antagonistic ability index;

[0027] Based on the functional trait table of the core strains, statistical calculations were performed, and the 25th and 75th percentiles were used as the thresholds for differential grading. Strategy classification logic was set to divide resource intake levels and antagonistic levels, and the corresponding matching strategies were combined and invoked to construct a set of differentiated competitive strategies.

[0028] Specifically, the steps for calculating the revenue matrix library also include:

[0029] For any two participants , Call the corresponding resource intake index , Combined with resource return coefficients, the participants are calculated. , Resource competition benefits , ;

[0030] Synchronous call participants , Antagonistic ability index , Combined with the antagonistic payoff coefficient, the participants were calculated. , Antagonistic ability benefit , ;

[0031] Calculate the total revenue of participants, construct a binary revenue matrix, traverse all core strain combinations, and generate a revenue matrix library.

[0032] Based on the predicted scenario sequence, the temperature correction factor and humidity correction factor for each scenario are calculated, thereby calculating the environmental correction factor;

[0033] Set upper and lower limits for environmental impact, and adjust the environmental correction factor accordingly;

[0034] Based on the corrected environmental correction coefficient, the total revenue is corrected to obtain the corrected binary revenue matrix, thereby generating the corrected revenue matrix library.

[0035] Specifically, the steps for solving scenario-based game theory include:

[0036] Set abundance constraints so that the equilibrium abundance of each participant must not exceed the constraint multiple of the initial relative abundance;

[0037] For each scenario, load the modified binary payoff matrix for all participants, and randomly assign an initial strategy to each participant, calculating the initial payoff based on the initial relative abundance.

[0038] Iterate through each participant, switch to other strategies, and calculate the switching payoff;

[0039] If the switching benefit is greater than the initial benefit and the abundance after the switching does not exceed the abundance constraint, then the new strategy is retained; otherwise, the original strategy is maintained.

[0040] When all participants have not changed their strategies in multiple consecutive iterations, it is determined that an equilibrium state has been reached, the iteration stops, and a sub-scenario equilibrium result table is generated for each scenario.

[0041] Specifically, the steps for solving scenario-based games also include:

[0042] For each scenario's sub-scenario equilibrium result table, sort the equilibrium abundance in descending order and select the top results with the highest equilibrium abundance in each scenario. Each bacterial species is defined as the dominant bacterial species in the corresponding scenario;

[0043] The dominant strains in all scenarios are integrated to generate a balanced summary table;

[0044] The participants were grouped according to scenario type, and the average equilibrium abundance of each participant in each scenario group was calculated.

[0045] Specifically, the steps for finding the optimal solution include:

[0046] Define state input, environmental input, and regulation input, and combine the logic of scenario-based game solving to construct a regulation response prediction model to output the equilibrium state of the microbial community after regulation.

[0047] For each scenario, multiple sets of regulatory input vectors with different intensities are constructed, sample labels are calculated, training samples are generated, and training and testing sets are divided to train the regulatory response prediction model.

[0048] By setting robustness constraints and embedding expert rules, a multi-objective evolutionary algorithm is used to generate an initial set of candidate solutions, with the control parameter vector as the decision variable.

[0049] Specifically, the steps for finding the optimal solution also include:

[0050] For each scenario, each set of candidate solutions is input into the regulation response prediction model, and combined with robustness constraints, an initial evaluation table is formed.

[0051] Using the candidate schemes in the initial evaluation table as the parent population, offspring schemes are generated through crossover and mutation operations, combined with robustness evaluation.

[0052] The parent and child schemes are merged, and non-dominated schemes are selected according to the non-dominated sorting rules. The non-dominated schemes are used as the next generation of parent population. After multiple rounds of iteration, a set of non-dominated schemes is generated, and the non-dominated schemes are sorted in descending order.

[0053] The soil microbial community regulation system includes: an analysis module, a competitive game module, and an optimization module;

[0054] The analysis module is used to acquire multidimensional data of the target soil area, generate multiple environmental scenario sequences by combining meteorological forecasts, and screen core microbial species.

[0055] The competitive game module is used to construct differentiated competitive strategies and a payoff matrix library, combine scenario correction coefficients to perform scenario-based corrections on the payoff matrix, and solve the community equilibrium state for each scenario.

[0056] The optimization module is used to construct a regulation response prediction model, set multiple objectives, and select the optimal regulation scheme by combining ecological security thresholds and expert rules.

[0057] The beneficial effects of this invention are:

[0058] By combining sensor and sampling technologies to acquire environmental data, and then using high-throughput sequencing and metagenomic analysis to screen core bacterial species, extract functional traits such as resource uptake and antagonistic inhibition, the participants and capabilities in interspecies competition are clearly defined. This avoids distortion of competition logic due to the omission of core bacterial species or functional ambiguity. Based on game theory, a scenario-based competition model is constructed, transforming bacterial interactions into quantitative strategies and payoff matrices. The competitive payoff is dynamically adjusted by incorporating environmental correction coefficients, solving the problems of neglecting resource competition and antagonistic inhibition interactions. This makes the prediction of microbial community evolution more consistent with actual laws and corrects the prediction bias of traditional models. On this basis, a regulatory response prediction model is constructed, embedding game logic to accurately output the microbial community state after regulation, avoiding the deviation of regulatory measures from the actual competition. Multi-objective robust optimization is used to screen the Pareto optimal solution through multi-objective setting and full-scenario constraints. This ensures that the regulatory measures can achieve the expected effect by guiding interspecies competition, while avoiding the disruption of the original microbial community balance, and also takes into account economic feasibility. This solves the problem of regulatory failure caused by neglecting interspecies interactions in traditional solutions, achieving a unity of scientificity, effectiveness, and safety in soil microecological regulation, and the solution is adaptable to different environmental fluctuations. Attached Figure Description

[0059] Figure 1 A schematic diagram illustrating methods for regulating soil microbial communities;

[0060] Figure 2 This is a flowchart of the process for screening core bacterial strains in this invention;

[0061] Figure 3This is a flowchart of the scenario-based game solving process in this invention;

[0062] Figure 4 This is a flowchart for finding the optimal solution in this invention. Detailed Implementation

[0063] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0064] Example 1

[0065] refer to Figures 1 to 4 As shown in the figure, this embodiment introduces a method for regulating soil microbial communities, including the following steps:

[0066] Based on the geographical location, planting type, and plot boundaries of the target soil area, a soil sensor array, such as temperature, humidity, and pH sensors, is deployed to continuously acquire real-time environmental data at a preset frequency. This accumulates historical climate data for the target soil area. Simultaneously, historical agricultural intervention records for the target soil area are acquired, including fertilizer type and dosage, microbial inoculation time, and irrigation frequency. These records are integrated along a timeline to construct an intervention dataset containing the correspondence between intervention measures and their implementation times. Combined with future climate forecast data released by meteorological departments, using temperature and humidity as key scenario factors, Monte Carlo simulations are used to generate predicted scenario sequences. The expected scenario sequence includes various environmental fluctuation scenarios, such as normal, drought and high temperature, and rainy and cold. For the number of scenario types, For the first This scenario ,scene Dynamic curves showing temperature and humidity changes over time , To cover the impact of environmental uncertainty on interspecific competition;

[0067] Since interspecific competition is essentially about the interaction between different microorganisms, this study identifies soil microbial species, such as highly abundant beneficial and pathogenic bacteria, to avoid overlooking key competitors. Based on a pre-set environmental sampling depth, soil samples from the target area are collected using a five-point sampling method and a grid-based approach. DNA is extracted, and high-throughput sequencing is performed using pre-set quality control standards. The sequencing data is then assembled and annotated to generate a microbial community composition table, including the species names of all microorganisms to be tested within the target soil area. The relative abundance of each species is calculated, and the species are ranked based on their relative abundance to identify core microbial species and generate a core microbial species list. This provides a target for extracting functional traits. Based on the gene sequence information of the core bacterial species, metagenomic sequencing is performed to provide a quantitative basis for interspecific competition modeling. Through comparison with a pre-set functional database, functional trait data of the core bacterial species are extracted to form a functional trait table of the core bacterial species, which includes key functional information of each core bacterial species, such as resource uptake data, antagonistic inhibition data, and nutrient transformation data. Resource uptake data includes the abundance of each coding gene to reflect resource utilization efficiency. Antagonistic inhibition data includes the expression level of antibiotic synthesis genes. Nutrient transformation data includes the presence or absence of nitrogen fixation genes and phosphorus solubilization genes to reflect functional value.

[0068] Low-abundance microbial species have minimal impact on the overall community structure; therefore, engaging in game theory with them would increase computational complexity and be impractical. Instead, core microbial species are used as game participants, each assigned a unique identifier. Since interspecies competition follows a non-cooperative principle, each species prioritizes maximizing its own population growth without proactive collaboration. Based on the ecological competition hypothesis, a set of participant strategies is constructed, and multiple differentiated competitive strategies are developed for each core microbial species. These strategies are correlated with the functional traits of the core species to ensure that the strategy design aligns with actual microbial competitive behavior, such as resource competition and antagonistic inhibition, rather than subjective assumptions. For any two participants, the strategy set serves as the framework. A binary payoff matrix is ​​constructed to generate a payoff matrix library for multi-species interactions, enabling the quantitative transformation of competitive advantage. Since the competitive ability of microorganisms is significantly affected by the environment, an environmental correction coefficient is set for each scenario based on the expected scenario sequence to correct the payoff matrix. An improved Nash equilibrium solution algorithm is used to solve the game in different scenarios. For each scenario, the core microbial structure in the equilibrium state is obtained, including the optimal strategy selection and equilibrium abundance of each core microbial species. The equilibrium results under all scenarios are integrated, and the dominant microbial species under each scenario are labeled to generate an equilibrium summary table to truly reflect the natural stable state of the microbial community without regulation.

[0069] By combining scenario-based game theory solutions, a regulatory response prediction model is constructed to predict the state of the microbial community after regulation. A multi-objective robust optimization problem is defined, setting multiple optimization objectives: maximizing the fit between the regulated microbial community and the ideal state, minimizing the disturbance to the indigenous microbial community, and minimizing the regulation cost. Ecological safety threshold constraints and expert rules are introduced under all scenarios. A multi-objective evolutionary algorithm combining scenario-based robust optimization is used to solve the problem, generating candidate regulation schemes. The regulatory response prediction model is used to predict and evaluate the candidate regulation schemes, and the optimal scheme is solved to achieve the optimal trade-off between effectiveness, safety, and economy.

[0070] Specifically, the steps for screening core bacterial strains include:

[0071] A secondary classification threshold was set, and combined with organic matter content data, the target soil area was divided into sampling zones: high-trophic, mesotrophic, and low-trophic zones. This ensured that the samples covered micro-domains with different resource abundances, avoiding insufficient representativeness of the microbial community due to sample homogeneity. Within each type of sampling zone, grids were divided according to preset specifications. A five-point sampling method was used to determine the sampling location within each grid, including multiple sub-sampling points. The coordinates of each sampling point were recorded using GPS, and a sampling point distribution map was constructed. The sampling zone type and the location of the sampling points were marked on the distribution map to facilitate subsequent traceability of the sample source. Organic content data was obtained by using sensors as the initial collection points.

[0072] The environmental sampling depth is set, and soil is collected using a sterile sampler. Multiple subsamples are collected at each sampling point to reduce the random error of single-point sampling. Soil from multiple subsample points within the same grid is mixed, and impurities are removed. The samples are then placed into sterile sampling bags with sample numbers and labeled with sampling area type, sampling point number, and time. All mixed samples are integrated to generate a soil sample set. Each soil sample contains sample number, sampling area type, coordinates, and collection time.

[0073] Total DNA was extracted from soil samples using the CTAB method, and the purity and integrity of the DNA were tested. Combined with high-throughput sequencing, the DNA was broken down into a large number of short gene fragments using sequencing instruments to obtain the raw sequence data of the soil samples. Low-quality data was filtered using quality control standards, and sequence splicing, redundancy removal, and OTU clustering were performed. Based on species classification databases, such as the Silva database and the UNITE database, species classification was performed, and the data was integrated based on each soil sample to generate a microbial community composition table. At this point, sequencing was only performed at the community structure level to quickly obtain the species composition and relative abundance of all microorganisms in the soil. Only conserved genes (such as 16S rRNA) were sequenced, and the specific functional genes of the bacterial species were not analyzed.

[0074] Based on the microbial community composition table, the relative abundance of each species is calculated, and the species are sorted in descending order based on the relative abundance to generate a species abundance ranking table, which includes species name, relative abundance, and taxonomic group (beneficial bacteria / pathogens).

[0075] Set a core abundance threshold, accumulate the relative abundance of species after sorting, stop accumulating once the accumulated relative abundance exceeds the core abundance threshold, and define all species included in the accumulation as core species, thereby generating a core species list, including core species number, species name, relative abundance, and functional type;

[0076] Based on each soil sample, the relative abundance of each core microbial species is extracted, and an effective abundance threshold is set to determine whether the core microbial species in a single soil sample is effective. If the relative abundance is less than the effective abundance threshold, the core microbial species in the corresponding soil sample is determined to be an ineffective core microbial species. If the relative abundance is greater than or equal to the effective abundance threshold, the core microbial species in the corresponding soil sample is determined to be an effective core microbial species.

[0077] The number of effective core microbial species in a single soil sample is counted, and the core microbial species coverage rate of a single soil sample is calculated by comparing the number of effective core microbial species in the single soil sample with the total number of core microbial species in the single soil sample.

[0078] Set an effective coverage threshold and screen soil samples with a core microbial species coverage rate not less than the effective coverage threshold. These samples are defined as qualified samples to ensure that the genetic information of the core microbial species is complete. Since the core microbial species functions differ in different trophic zones, if the qualified samples do not cover each type of sampling zone, soil samples from the uncovered sampling zones are collected to ensure that there is one qualified sample in each of the high-trophic zone, medium-trophic zone, and low-trophic zone. Finally, the zonal sequencing samples are generated.

[0079] Metagenomic sequencing was performed on the generated partitioned sequencing samples. DNA was extracted from each qualified sample in the partitioned sequencing samples, and the DNA was broken into multiple fragments using an ultrasonic disruptor. End repair, A-tailing, and sequencing adapter ligation were performed sequentially. Sequencing libraries were constructed by PCR amplification, and the library quality was tested using a bioanalyzer. Metagenomic sequencing was performed on the sequencing libraries that passed the quality test to obtain raw sequence data. Data filtering was performed to obtain clean data for each qualified sample. Genome assembly was performed on the clean data of each qualified sample, and ORF prediction was performed. An independent ORF prediction result was generated for each qualified sample, thereby generating the ORF sequence of each qualified sample. The reference genome of each core bacterial species was obtained. The ORF sequences of all qualified samples were compared with the reference genome to screen out the ORF sequences corresponding to the core bacterial species. The ORF sequences of the core bacterial species of each qualified sample were merged, and redundancy was removed using CD-HIT software to obtain the gene sequence library of the core bacterial species, which contains the non-redundant functional gene sequences of all core bacterial species, and the gene ID and the name of the core bacterial species to which they belong are labeled.

[0080] Gene sequences obtained from metagenomic sequencing are compared with functional databases to annotate the functional genes of core bacterial species and generate a functional trait table of core bacterial species, including the name of the core bacterial species, resource uptake data, antagonistic inhibition data, and nutrient transformation data.

[0081] Specifically, the steps for solving scenario-based game theory problems include:

[0082] Since low-abundance microbial species account for a very low proportion of biomass in soil microecological communities and have a negligible impact on the overall community structure, core microbial species are used as game participants. A unique game participant identifier is assigned to each core microbial species. The participant identifier is associated with the number of the core microbial species, and the participant identifier is bound to the basic information of the core microbial species to form a game basic table to avoid confusion of species with the same name or data mismatch with species.

[0083] Based on organic matter content, a resource weight coefficient is set. Based on the functional trait table of the core strain, the abundance of each coding gene in the resource uptake data is averaged and summed, and then multiplied by the resource weight coefficient to calculate the resource uptake index.

[0084] Since the core strain is subjected to scenario-based game theory through resource uptake ability and antagonistic inhibition ability, the sum of the resource weight coefficient and the antagonistic weight coefficient is set to 1. Based on the resource weight coefficient, the antagonistic weight coefficient is calculated, the antagonistic inhibition data is averaged and summed, and then multiplied by the antagonistic weight coefficient to calculate the antagonistic ability index.

[0085] Based on the functional trait table of the core strain, statistical calculations were performed, and the 25th and 75th percentiles were used as the thresholds for difference classification. A strategy classification logic was established. For the resource uptake index, the 25th and 75th percentiles obtained through statistical calculations were used to set the primary and secondary resource difference thresholds, classifying the resource uptake index into high, medium, and low resource uptake levels. The primary resource difference threshold was less than the secondary resource difference threshold. For the antagonistic ability index, the 25th and 75th percentiles obtained through statistical calculations were used to set the primary and secondary antagonistic ability thresholds, classifying the antagonistic ability index into high, medium, and low antagonism levels. The primary antagonistic ability threshold was less than the secondary antagonistic ability threshold. Based on the real-time calculated resource uptake index and antagonistic ability index, each participant's resource uptake ability and antagonistic ability were classified into different levels.

[0086] Different resource uptake levels and antagonism levels are combined, and corresponding matching strategies are called from the strategy library. The functional data basis of the strategy is noted to form a differentiated competitive strategy set for each participant, such as high resource uptake + medium antagonism, medium resource uptake + high antagonism, to reflect the competitive behavior tendencies in the actual soil environment.

[0087] The relative abundance growth rate of microbial species is defined as the benefit indicator to reflect the competitive advantage of soil microorganisms. A positive abundance growth rate indicates that the strategy has helped the microorganisms gain a competitive advantage and the abundance has increased. A negative abundance growth rate indicates that the strategy is not conducive to the competition of microorganisms and the abundance has decreased. A reasonable benefit range is set in combination with the natural fluctuation range of microbial abundance.

[0088] For any two participants , Call the corresponding resource intake index , ,calculate and Exponential difference, that is Combined with resource return coefficients, the participants are calculated. Resource competition benefits Because resource competition is a zero-sum game, the participants The benefits are for the participants The losses for the participants Resource competition benefits ,in, , The resource yield coefficient represents the average abundance growth rate of the core bacterial species, where the number of participants in the game is represented by the number of participants.

[0089] Consistent with the calculation of resource competition benefits, participants are called synchronously. , Antagonistic ability index , Combined with the antagonistic payoff coefficient, the participants were calculated. , Antagonistic ability benefit , The total payout for each participant is calculated by summing the gains from resource contention and the gains from antagonistic capabilities. The strategy set is used as the rows of the matrix and the participants. The strategy set is used as the column of the matrix, and the total return is filled into the corresponding symmetrical position to construct a binary return matrix. The rationality of the total return of all strategy combinations is checked. If the total return is not within the reasonable range, the corresponding return coefficient is adjusted to recalculate the total return until the total return is within the reasonable range to avoid extreme returns. Among them, the antagonistic return coefficient is the correlation between the expression level and abundance of the antagonistic gene of the core bacterial species.

[0090] Iterate through all core strain combinations, generate multiple binary payoff matrices, integrate them to create a payoff matrix library, and then categorize them by participant. Identifier - Participant The matrix is ​​named according to the format of the identifier, and the original data for the revenue calculation is noted in the revenue matrix library;

[0091] Based on the predicted scenario sequence, the classic Q10 model in microbial ecology was adopted. Temperature correction coefficients were calculated for each scenario based on the difference between the average temperature of the corresponding scenario and the optimal growth temperature of microorganisms, combined with the temperature coefficient. Humidity correction coefficients were calculated for each scenario based on the ratio of the average humidity of the corresponding scenario to the suitable humidity for microorganisms. The temperature coefficient is the conventional Q10 value of soil microorganisms, calibrated with reference to historical climate data. The closer the temperature is to the optimal temperature, the larger the temperature correction coefficient, indicating stronger microbial competitiveness. The suitable humidity for microorganisms was calculated based on the humidity distribution of high-abundance microbial species; the closer the humidity is to the suitable humidity, the larger the humidity correction coefficient.

[0092] The temperature correction factor and the humidity correction factor are multiplied to obtain the environmental correction factor for each scenario. An upper limit and a lower limit of environmental impact are set. When the environmental correction factor is less than the lower limit of environmental impact, the environmental correction factor is corrected to the lower limit of environmental impact. When the environmental correction factor is greater than the upper limit of environmental impact, the environmental correction factor is corrected to the upper limit of environmental impact. This avoids extreme environments from causing the benefit value to exceed the actual physiological range of microbial growth.

[0093] Traverse the binary payoff matrix in the payoff matrix library. For each scenario, multiply all total payoffs in the matrix by the corresponding scenario's environmental correction coefficient to obtain the corrected total payoff. If the corrected total payoff exceeds the reasonable payoff range, adjust the environmental correction coefficient to obtain the corrected binary payoff matrix. Generate the corrected payoff matrix library to ensure that the payoff values ​​conform to the actual limitations of microbial growth, so that the subsequent game results truly reflect the competitive situation of the microbial community under different environments.

[0094] Traditional Nash equilibrium algorithms rely solely on the assumption that participants cannot improve their own gains by unilaterally changing their strategies. However, in soil microbial game theory, there is a problem where the equilibrium abundance far exceeds actual physiological limits. For example, a certain microbial species may have a very low initial relative abundance, but due to high strategy gains, an extremely high equilibrium abundance is predicted, which does not conform to the natural growth patterns of microorganisms. To address this, an abundance constraint is introduced: the equilibrium abundance of each participant must not exceed a constraint multiple of the initial relative abundance to avoid high strategy gains that are actually unattainable. Here, the initial relative abundance is the relative abundance in the core microbial species list, and the constraint multiple is set based on the upper limit of the natural growth of microbial abundance. Based on the equilibrium gains obtained from the equilibrium solution, the equilibrium coefficient is obtained by adding a base coefficient to the ratio of the equilibrium gain to the gain coefficient. The equilibrium abundance is then obtained by multiplying the equilibrium coefficient by the initial relative abundance. In this embodiment, the base coefficient is set to 1.

[0095] During the solution process, for each scenario, the modified binary payoff matrix of all participants is loaded, and an initial strategy is randomly assigned to each participant. The initial payoff is calculated based on the initial relative abundance. Each participant is iterated through, and other strategies are switched to calculate the switching payoff. If the switching payoff is greater than the initial payoff and the abundance after the switch does not exceed the abundance constraint, the new strategy is retained; otherwise, the original strategy is maintained.

[0096] When all participants do not change their strategies in multiple consecutive iterations, an equilibrium state is reached, and iteration stops. At the same time, an upper limit on the number of iterations is set to avoid infinite loops. At this point, the optimal strategy choice and corresponding equilibrium abundance of each participant are recorded to complete the equilibrium solution for each scenario and generate a sub-scenario equilibrium result table for each scenario, including scenario number, participant identifier, optimal strategy, equilibrium benefit, and equilibrium abundance. This satisfies the requirement of optimal strategy benefit while conforming to the actual growth limitations of microorganisms, and avoids extreme abundances that are out of touch with reality, so as to reflect the natural stability trend of the microbial community in this environment. At the same time, the rationality of the sub-scenario equilibrium result table for each scenario is checked. If the equilibrium abundance of a participant is less than the extinction threshold, its strategy combination and benefit calculation are backtracked to confirm whether it is due to the low antagonistic benefit, such as the suppression of a high antagonistic strategy by a low antagonistic strategy, to avoid algorithm error.

[0097] For each scenario's sub-scenario equilibrium result table, sort the equilibrium abundance in descending order and select the top results with the highest equilibrium abundance in each scenario. Each bacterial species is defined as the dominant bacterial species in the corresponding scenario. The functional characteristics of the dominant bacterial species are extracted from the functional trait table of the core bacterial species to clarify the dominant bacterial community types under different scenarios and construct the scenario dominant bacterial species table for each scenario.

[0098] The dominant microbial species tables for all scenarios are integrated to generate a balanced summary table, which includes scenario number, scenario type, optimal strategy of all participants, balanced abundance of all participants, dominant microbial species, and functional characteristics of dominant microbial species. The scenarios are grouped by scenario type, and the average balanced abundance of each participant in each scenario group is calculated. The scenario type and average abundance are labeled to intuitively present the overall impact of different environmental types on the microbial community structure.

[0099] Specifically, the steps for finding the optimal solution include:

[0100] The model calls the initial bacterial abundance and functional trait data of the core bacterial species under no regulation to ensure that the initial state of the model is consistent with the starting point of the competition of the real bacterial community. This is defined as the state input. The model calls the expected scenario sequence and the environmental correction coefficient of the corresponding scenario. This is defined as the environmental input, so that the model can perceive the impact of different environments on the competition. The model calls the parameters of the measures that affect the bacterial community in the intervention dataset and converts them into vectors. This is defined as the regulation input vector. Combined with the logic of solving the game in different scenarios, a regulation response prediction model is constructed to output the equilibrium state of the bacterial community after regulation, including the equilibrium abundance of the core bacterial species and the optimal competition strategy. This facilitates subsequent comparison with the unregulated state and allows for intuitive judgment of the regulation effect.

[0101] For each scenario, multiple sets of regulatory input vectors with different intensities are constructed to form sample combination pairs containing the initial microbial community, environmental scenario, and regulatory parameters. An improved Nash equilibrium algorithm is called to calculate the microbial community equilibrium state after applying regulatory parameters to each sample in real time as the sample label. This avoids the limitation that pre-stored labels cannot cope with dynamic regulation and ensures that the labels truly reflect the changes in the microbial community after interspecies competition, thereby generating training samples.

[0102] The training samples are divided into training and test sets, ensuring that each scenario is distributed in both sets. To simultaneously focus on abundance prediction accuracy and policy matching accuracy, the weighted sum of abundance error and matching policy error is used as the loss error, and the stochastic gradient descent algorithm is employed to minimize the loss. Meanwhile, extreme scenarios are selected from the predicted scenario sequence, and multiple sets of control parameters are added to generate supplementary samples. The Optuna library is used to perform sensitivity analysis on the learning rate and the number of hidden layer nodes, selecting the optimal hyperparameter combination to ensure that the control response prediction model maintains high prediction accuracy even under extreme scenarios, avoiding the problem of control failure caused by sudden weather in agricultural production. Finally, the trained control response prediction model is output.

[0103] Multiple optimization objectives were set, using the microbial community status of healthy soil as the baseline. The Bray-Curtis similarity coefficient was used to quantify the fit between the regulated microbial community and the ideal state. Maximizing this fit was the effectiveness objective; a coefficient closer to 1 indicated better regulation, ensuring the scheme effectively improved microbial health and increased abundance compared to the unregulated state. Non-pathogenic microbial species are defined as indigenous microbial communities. The disturbance to indigenous microbial communities is quantified by the Bray-Curtis dissimilarity index. At the same time, ecological safety thresholds are set based on soil microecological standards, such as microbial community diversity and lower limits of key functional bacteria abundance, to avoid disrupting the original microecological balance. Under the limit of ecological safety thresholds, the safety goal is to minimize disturbance. The weighted sum of the costs of various control measures is defined as the control cost. The goal is to minimize the control cost, ensuring that the plan is within the economic affordability of farmers or enterprises.

[0104] Robust constraints are set so that the microbial community state predicted by the control response prediction model meets the ecological safety threshold in all scenarios, avoiding the situation where it is effective in normal scenarios but fails in extreme scenarios. For example, a certain scheme performs well under normal weather conditions, but pathogens exceed the standard during drought.

[0105] Simultaneously embed expert rules and extract hard constraint rules from the soil microbial regulation knowledge base, such as inoculating with specific antagonistic bacteria exceeding the minimum dosage when the predicted pathogen abundance is too high, and prohibiting the use of acidic bacterial agents when the soil pH is too low, so as to eliminate ecological risks not identified by the regulation response prediction model and generate a list of expert rules.

[0106] Using a multi-objective evolutionary algorithm, such as NSGA-III, the regulation parameter vector is used as the decision variable. Within the range of the decision variable's values, multiple initial candidate schemes are randomly generated using the Latin hypercube sampling method to ensure coverage of different regulation intensities and combinations. At the same time, the list of prior expert rules is called to perform the first round of filtering on the multiple initial candidate schemes to generate an initial candidate scheme set.

[0107] Robustness assessment is performed on each candidate scheme. Each scenario is loaded sequentially, and each group of candidate schemes is input into the regulation response prediction model to calculate regulation measures, game correction, and equilibrium prediction. This yields the microbial community equilibrium state of the candidate scheme under different scenarios, and the objective function value under each scenario is calculated, including fit, perturbation, and regulation cost. The worst performance of each objective is taken as the evaluation result of the scheme to effectively avoid extreme environmental risks. For example, a certain scheme may be low-cost and effective in most scenarios, but may cause excessive disturbance to the indigenous microbial community in rainy scenarios. If any item of the evaluation result does not meet the robustness constraint, the corresponding candidate scheme is removed, forming an initial evaluation table.

[0108] Using the candidate schemes in the initial evaluation table as the parent population, simulated binary crossover is used to randomly select two parent schemes and generate two offspring in each decision variable dimension. This ensures that the offspring parameters are within the range of the parent values ​​and conform to the regulation rules. At the same time, multinomial mutation is used to make small adjustments to the parameters of each offspring scheme. The mutation amplitude is controlled within the mutation ratio threshold of the parameter value range to avoid generating schemes that exceed the safe range. This generates a new generation of candidate schemes, and a robustness evaluation is performed to select offspring schemes that pass the robustness evaluation.

[0109] The parent and child schemes are merged, and non-dominated schemes are selected according to the non-dominated sorting rule. Schemes that are better than other schemes in all objectives are retained. The selected non-dominated schemes are used as the next generation of parent population. After multiple rounds of iteration, a set of non-dominated schemes is generated.

[0110] The set of non-dominated schemes in the final round is sorted in descending order of effectiveness, and a Pareto optimal frontier is plotted. Each point on the frontier represents a control scheme that achieves the best balance among effectiveness, safety, and economy. The control parameters and performance of each target are clearly marked.

[0111] Example 2

[0112] Another embodiment of the present invention provides a soil microecological microbial community regulation system, comprising: an analysis module, a competitive game module, and an optimization module;

[0113] The analysis module is used to acquire environmental data, historical agricultural intervention records, and soil samples of the target soil area in real time. At the same time, it combines meteorological forecasts to generate multiple environmental scenario sequences to ensure that the subsequent analysis is supported by comprehensive data without single bias, avoiding the regulatory logic gap caused by missing data. It also performs DNA extraction, high-throughput sequencing, and metagenomic analysis on soil samples to generate a microbial community composition table. Based on relative abundance, core species are screened and functional trait data are extracted to ensure that the subsequent competitive modeling has clear participants and capabilities, avoiding the distortion of game logic caused by the ambiguity of species functions.

[0114] The competitive game module defines core microbial species as game participants. Based on the resource uptake index and antagonistic ability index of core microbial species, it constructs differentiated competitive strategies to ensure that the strategies match the actual competitive capabilities of the species. Using the relative abundance growth rate of the species as the payoff indicator, it constructs a payoff matrix library and combines scenario correction coefficients to perform scenario-based corrections to avoid payoff quantification deviations caused by environmental fluctuations. An improved Nash equilibrium solution algorithm is used to solve the microbial community equilibrium state under different scenarios, outputting the optimal strategy and equilibrium abundance of the core microbial species under each scenario. This accurately quantifies the impact of resource competition and antagonistic interactions on microbial community evolution, solves the prediction bias problem caused by neglecting interspecies competition in traditional schemes, and makes the prediction of microbial community stability highly consistent with the actual microecological laws of the soil.

[0115] The optimization module is used to embed the game theory model logic, build a regulation response prediction model, predict the balance state of the microbial community after regulation, ensure that the regulation effect is predicted in advance, avoid microbial imbalance caused by blind intervention, and set multiple objectives. Combining ecological safety thresholds and expert rules, it selects the Pareto optimal regulation scheme to ensure that the output scheme is safe and feasible under different environmental scenarios and balances the regulation effect with actual production needs.

[0116] Working principle and effects:

[0117] By employing soil sensors, five-point sampling, and high-throughput sequencing, environmental data, microbial community data, and core microbial species functional data of the target area were collected. Monte Carlo simulations were used to generate multi-environmental scenario sequences, avoiding the interference of sample uniformity and environmental fluctuations on subsequent analysis. This provided a comprehensive data foundation for competitive modeling. Core microbial species were screened, low-abundance ineffective species were excluded, and their competitive capabilities were quantified. A scenario-specific competitive model was constructed using game theory, with core microbial species designated as non-cooperative game participants. Resource competition and antagonistic inhibition strategies were defined based on functional data. A payoff matrix was constructed and dynamically adjusted using environmental correction coefficients, addressing the problem of traditional techniques neglecting interspecies interactions. This ensured that microbial community evolution predictions aligned with actual patterns, corrected prediction biases, and constructed a regulatory response model embedded with game logic. This model accurately outputs the microbial community state after regulation. Finally, through multi-objective robust optimization, the Pareto optimal solution was selected across all scenarios. This ensured that regulation guided interspecies competition to achieve the expected results while avoiding disruption of the indigenous microbial community balance and controlling costs, ultimately achieving scientific, effective, and safe soil microecological regulation.

[0118] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for regulating the microbial community in soil microecology, characterized in that, include: Based on the target soil region, real-time environmental data is continuously collected and combined with future climate prediction data to generate a predicted scenario sequence using Monte Carlo simulation. A table of microbial community composition was constructed, and core bacterial species were screened and metagenomic sequencing was performed. Using the core bacterial species as game participants, a set of participant strategies is constructed, a payoff matrix library is calculated, and an environmental correction coefficient is set for each scenario. The game is then solved in different scenarios to generate an equilibrium summary table. A regulation response prediction model is constructed, multiple optimization objectives are set, robust constraints and expert rules are introduced, candidate regulation schemes are generated, and the optimal scheme is solved. The steps for calculating the revenue matrix library include: Using the core bacterial species as game participants, assign participant identifiers to form a basic game table; Set resource weight coefficients, average and sum the resource intake data, and calculate the resource intake index; Calculate the antagonistic weighting coefficient, average and sum the antagonistic inhibition data, and calculate the antagonistic ability index; Based on the functional trait table of the core strains, statistical calculations were performed, and the 25th and 75th percentiles were taken as the thresholds for differential grading. Strategy classification logic was set to divide resource uptake levels and antagonistic levels, and the corresponding matching strategies were combined and invoked to construct a set of differentiated competitive strategies. For any two participants , Call the corresponding resource intake index , Combined with resource return coefficients, the participants are calculated. , Resource competition benefits , ; Synchronous call participants , Antagonistic ability index , Combined with the antagonistic payoff coefficient, the participants were calculated. , Antagonistic ability benefit , ; Calculate the total revenue of participants, construct a binary revenue matrix, traverse all core strain combinations, and generate a revenue matrix library. Based on the predicted scenario sequence, the temperature correction factor and humidity correction factor for each scenario are calculated, thereby calculating the environmental correction factor; Set upper and lower limits for environmental impact, and adjust the environmental correction factor accordingly; Based on the corrected environmental correction coefficient, the total revenue is corrected to obtain the corrected binary revenue matrix, thereby generating the corrected revenue matrix library.

2. The method for regulating soil microbial communities according to claim 1, characterized in that, The steps for screening core strains include: Set a secondary division threshold to divide the target soil area into sampling zones and construct a sampling point distribution map; Soil sample sets are generated by collecting samples based on environmental sampling depth. DNA was extracted from soil samples and combined with high-throughput sequencing to obtain the raw sequence data of the soil samples. The data was then filtered according to quality control standards to generate a table of microbial community composition. Calculate the relative abundance of each species, sort them in descending order, and generate a species abundance ranking table. The relative abundances after sorting are summed to screen for core strains and form a list of core strains.

3. The method for regulating soil microbial communities according to claim 2, characterized in that, The steps for screening core strains also include: The relative abundance of each core microbial species in a single soil sample was extracted, and invalid core microbial species in a single soil sample were screened out by an effective abundance threshold. The number of invalid core microbial species in a single soil sample was counted, and the coverage rate of core microbial species was calculated. Based on the effective coverage threshold, qualified samples are screened to generate partitioned sequencing samples; Based on functional database comparison, a functional trait table of core strains is generated, including resource uptake data, antagonistic inhibition data, and nutrient transformation data.

4. The method for regulating soil microbial communities according to claim 3, characterized in that, The steps for solving scenario-based game theory include: Set abundance constraints so that the equilibrium abundance of each participant must not exceed the constraint multiple of the initial relative abundance; For each scenario, load the modified binary payoff matrix for all participants, and randomly assign an initial strategy to each participant, calculating the initial payoff based on the initial relative abundance. Iterate through each participant, switch to other strategies, and calculate the switching payoff; If the switching benefit is greater than the initial benefit and the abundance after the switching does not exceed the abundance constraint, then the new strategy is retained; otherwise, the original strategy is maintained. When all participants have not changed their strategies in multiple consecutive iterations, it is determined that an equilibrium state has been reached, the iteration stops, and a sub-scenario equilibrium result table is generated for each scenario.

5. The method for regulating soil microbial communities according to claim 4, characterized in that, The steps for solving scenario-based game theory also include: For each scenario's sub-scenario equilibrium result table, sort the equilibrium abundance in descending order and select the top results with the highest equilibrium abundance in each scenario. Each bacterial species is defined as the dominant bacterial species in the corresponding scenario; The dominant strains in all scenarios are integrated to generate a balanced summary table; The participants were grouped according to scenario type, and the average equilibrium abundance of each participant in each scenario group was calculated.

6. The method for regulating soil microbial communities according to claim 5, characterized in that, The steps to find the optimal solution include: Define state input, environmental input, and regulation input, and combine the logic of scenario-based game solving to construct a regulation response prediction model to output the equilibrium state of the microbial community after regulation. For each scenario, multiple sets of regulatory input vectors with different intensities are constructed, sample labels are calculated, training samples are generated, and training and testing sets are divided to train the regulatory response prediction model. By setting robustness constraints and embedding expert rules, a multi-objective evolutionary algorithm is used to generate an initial set of candidate solutions, with the control parameter vector as the decision variable.

7. The method for regulating soil microbial communities according to claim 6, characterized in that, The steps for finding the optimal solution also include: For each scenario, each set of candidate solutions is input into the regulation response prediction model, and combined with robustness constraints, an initial evaluation table is formed. Using the candidate schemes in the initial evaluation table as the parent population, offspring schemes are generated through crossover and mutation operations, combined with robustness evaluation. The parent and child schemes are merged, and non-dominated schemes are selected according to the non-dominated sorting rules. The non-dominated schemes are used as the next generation of parent population. After multiple rounds of iteration, a set of non-dominated schemes is generated, and the non-dominated schemes are sorted in descending order.

8. A soil microbial community regulation system, used to implement the soil microbial community regulation method as described in any one of claims 1-7, characterized in that, include: Analysis module, competitive game module, and optimization module; The analysis module is used to acquire multidimensional data of the target soil area, generate multiple environmental scenario sequences by combining meteorological forecasts, and screen core microbial species. The competitive game module is used to construct differentiated competitive strategies and a payoff matrix library, combine scenario correction coefficients to perform scenario-based corrections on the payoff matrix, and solve the community equilibrium state for each scenario. The optimization module is used to construct a regulation response prediction model, set multiple objectives, and select the optimal regulation scheme by combining ecological security thresholds and expert rules.

Images