Intestinal flora colonization effect prediction and nutrient recommendation system based on knowledge graph

Abstract

This invention relates to the field of knowledge graph technology, specifically disclosing a knowledge graph-based system for predicting gut microbiota colonization effects and recommending nutrients. The system includes: knowledge graph construction, data acquisition, microbiota game modeling, dynamic competitive evolutionary reasoning, Nash equilibrium calculation, colonization probability prediction, and a strategic nutrient recommendation output device. It identifies microbiota interactions by constructing a heterogeneous semantic network, describes the strategic change paths of the microbiota under resource conditions using evolutionary game theory algorithms, and identifies the system's stable state and risks by combining Nash equilibrium points. By introducing a dynamic competitive network based on evolutionary game theory, this invention can accurately predict the outcome of microbiota competition, avoid resource competition risks, guide the gut towards a stable state dominated by beneficial bacteria, and improve the accuracy, safety, and regulatory effect of nutritional intervention.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology, specifically relating to a knowledge graph-based system for predicting gut microbiota colonization and recommending nutrients. Background Technology

[0002] With the rapid development of precision medicine and the big health industry, the regulation of the gut microbiota has become a key entry point for maintaining human health and preventing chronic diseases. The gut microbiota directly affects the host's immune regulation, nutrient metabolism, and nervous system function through its metabolites, and the balance of its composition plays a vital role in human physiological health. Under the trend of digital health management, utilizing big data and information technology for precise modeling and intervention of the gut microbiota is gradually becoming a core research direction in the field of personalized nutrition.

[0003] Knowledge graph-based microbiome interaction analysis and nutrient recommendation technology is a crucial pathway for achieving precise regulation of the gut microbiota. This technology aims to quantify the impact of different exogenous interventions on microbiome structure by constructing a semantic network encompassing bacterial species, metabolites, nutrients, and their relationships, and based on this, provide users with customized dietary or supplementation plans. Through the structured integration of large-scale biomedical data, the system can assist researchers in analyzing complex microbial interactions, providing decision support for promoting the colonization of beneficial bacteria.

[0004] Current technologies for modeling gut microbiota relationships largely rely on static weight analysis, which struggles to accurately depict the complex competitive effects within the gut microbiota. Traditional recommendation models are typically based on unidirectional promotion logic, neglecting the collaborative competition for nutrient resources by harmful bacteria. This can lead to unexpected outbreaks of pathogenic bacteria or even the risk of colonization reversal after nutrient supplementation. Current analytical methods lack the ability to dynamically extrapolate ecological evolution processes, failing to identify microbiota strategy games in resource-constrained environments and struggling to pinpoint the balance of benefits between different microbiota, resulting in discrepancies between the accuracy of recommendation results and actual clinical outcomes. Summary of the Invention

[0005] The purpose of this invention is to provide a knowledge graph-based system for predicting gut microbiota colonization effects and recommending nutrients, which can effectively solve the problem of ineffective or even adverse nutrient supplementation caused by ignoring the community competition effect of microbiota in the above-mentioned background technology.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system includes a microecological knowledge graph construction device, a user gut sample data collection device, a microbiota game relationship modeling device, a dynamic competitive evolution reasoning device, a Nash equilibrium point calculation device, a colonization probability prediction device, and a strategic nutrient recommendation output device. The aforementioned microbial ecosystem knowledge graph construction device is used to construct a heterogeneous semantic network containing bacterial species nodes, metabolite nodes, nutrient nodes, and functional gene nodes by integrating biomedical literature, microbial community metabolism databases, and clinical nutrition research data. The device identifies the interaction relationships between different microbial communities through a preset entity extraction logic. These interaction relationships include mutualistic symbiosis, competitive inhibition, predation, and neutral metabolic associations. Furthermore, the device assigns biological attributes to each bacterial species node, including a growth rate constant, substrate affinity coefficient, and survival threshold under specific environmental conditions.

[0007] The user intestinal sample data acquisition device is used to acquire raw intestinal microbial sequencing data and clinical physiological indicators of the target user; the user intestinal sample data acquisition device performs standardized processing on the raw data to identify the distribution of dominant bacteria, abundance of potential pathogenic bacteria and bacterial diversity indicators in the user's intestine; the user intestinal sample data acquisition device converts the processed data into an initial state feature vector that the system can recognize, as the input benchmark for subsequent game theory inference.

[0008] The microbial community game relationship modeling device is used to map beneficial bacteria, harmful bacteria, and neutral bacteria in the gut to independent players in a game theory model based on the semantic network generated by the microecological domain knowledge graph construction device. The microbial community game relationship modeling device maps selectable nutrient combinations to resource acquisition strategies that players can choose, and transforms the colonization abundance, metabolite yield, and immune contribution of the microbial community under specific nutrient intervention into corresponding game payoff functions.

[0009] The dynamic competitive evolutionary reasoning device is used to simulate the dynamic confrontation process between players under different nutrient delivery strategies. The dynamic competitive evolutionary reasoning device breaks away from the traditional static weighted reasoning logic and introduces an evolutionary game algorithm to describe the strategy iteration path of the microbial community in a limited space and limited resource environment. The dynamic competitive evolutionary reasoning device is equipped with a replicated dynamic equation logic to describe the rate of change of the proportion of a specific microbial community in the competition process. By calculating the difference between the instantaneous gain of a specific strategy and the average gain of the community, the dynamic evolution of the intestinal ecosystem is deduced.

[0010] The Nash equilibrium point calculation device is used to perform convergence analysis on the evolutionary trajectory generated by the dynamic competitive evolutionary reasoning device to find the system stability point in a multi-party game environment. The Nash equilibrium point calculation device identifies the distribution of the microbial community system that eventually tends to be stable under specific nutrient intervention conditions through iterative calculation. The Nash equilibrium point calculation device is specially configured with risk identification logic to determine whether harmful bacteria have obtained a profit increase exceeding a preset threshold by seizing nutrient resources when reaching the equilibrium state, and to identify whether there is a risk of backlash after nutrient supplementation.

[0011] The colonization probability prediction device is used to quantify the colonization success rate of the target beneficial bacteria within the prediction period based on the output of the Nash equilibrium point calculation device; the colonization probability prediction device evaluates the degree of improvement of the current nutrient intervention strategy on the intestinal microecological balance by comparing the predicted equilibrium state with the preset healthy gut standard model.

[0012] The strategic nutrient recommendation output device is used to generate and output a personalized nutrient combination scheme for the target user when the colonization probability prediction device determines that the colonization success rate is higher than a preset probability threshold and the Nash equilibrium point calculation device determines that the system converges to an equilibrium state where beneficial bacteria dominate.

[0013] Preferably, the microbial ecosystem knowledge graph construction device uses natural language processing technology to extract the association logic between microbial communities and nutrients from unstructured text, and establishes multidimensional connections based on a preset ontology model.

[0014] Furthermore, the user intestinal sample data acquisition device includes a metagenomic data interface and a metabolomics data interface for receiving standardized documents from external testing institutions.

[0015] Furthermore, when defining the payoff function, the microbial community game relationship modeling device uses the substrate utilization rate of specific nutrients, the resistance to harmful bacterial secretions, and the adhesion ability to the intestinal mucosa as positive correction factors for the payoff, while using the environmental pressure coefficient and the metabolic waste inhibition effect as negative correction factors for the payoff.

[0016] Furthermore, the dynamic competitive evolutionary reasoning device constructs a dynamic competitive network to simulate in real time the topological changes of different bacterial community nodes during the resource grabbing process. These topological changes reflect the dynamic migration of the competitive intensity among bacterial communities.

[0017] Furthermore, the Nash equilibrium point calculation device adopts multi-objective optimization logic, which simultaneously considers maximizing the colonization abundance of beneficial bacteria, maximizing the growth inhibition rate of harmful bacteria, and minimizing the host metabolic load when searching for the equilibrium point.

[0018] Furthermore, the colonization probability prediction device is equipped with time-dimensional simulation logic to predict the persistence of microbial colonization and its robustness against external interference over different time spans.

[0019] Furthermore, the strategy nutrient recommendation output device generates a scheme that includes recommended types of prebiotics, specific amino acid ratios, trace element content, and suggested supplementation cycles.

[0020] Furthermore, the knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system also includes a feedback closed-loop optimization device. The feedback closed-loop optimization device acquires secondary gut microbiota test data after the user executes the recommendation plan and compares and analyzes it with the previous prediction results.

[0021] Furthermore, the feedback closed-loop optimization device automatically adjusts the weight of the payoff function in the microbial community game relationship modeling device and the evolution parameters in the dynamic competitive evolution inference device by calculating the prediction deviation.

[0022] Furthermore, when constructing the knowledge graph in the microbial ecosystem, the device refines the interactions between microbial communities into signal molecule-mediated communication, metabolite coupling, and steric hindrance competition, and assigns specific logical reasoning rules to each type of interaction.

[0023] Furthermore, the microbial community game relationship modeling device introduces the bounded rationality assumption when simulating player behavior, taking into account the predetermined delay in the response of microorganisms in complex environments, making the game model closer to the real biological evolution process.

[0024] Furthermore, the dynamic competitive evolutionary reasoning device uses matrix operation logic to process the pairwise game relationships between large-scale bacterial communities and summarizes them into an evolutionary trend map at the population level.

[0025] Furthermore, the Nash equilibrium point calculation device can identify multiple equilibrium states and, based on a preset optimal health standard, select the equilibrium point that provides the highest degree of improvement to the host's physiological indicators from multiple potential stable states.

[0026] Furthermore, the strategic nutrient recommendation output device automatically performs a safety screening before outputting a plan. If the prediction results show that the addition of a certain nutrient will cause harmful bacteria to achieve explosive growth through a specific metabolic pathway, the output of the plan will be immediately blocked, and the dynamic competitive evolutionary reasoning device will be called back to find an alternative strategy.

[0027] Furthermore, the system is also equipped with an environmental parameter correction device, which is used to perform personalized parameter calibration of the game payoff function based on the user's age, gender, region, and basal metabolic index.

[0028] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention successfully breaks through the limitations of static weight analysis in traditional microecological regulation by introducing a dynamic competition network based on evolutionary game theory. The system can accurately predict the dynamic competition results among bacterial communities, identify and avoid the resource grabbing risks common in traditional methods, and prevent unintended outbreaks of harmful bacteria due to misleading nutrient supplementation.

[0029] 2. By calculating the Nash equilibrium point of the intestinal ecosystem, this invention not only focuses on the successful colonization of a single population, but also ensures from a system-wide perspective that the recommended scheme can guide the intestinal flora structure towards a stable state where beneficial bacteria dominate, thereby improving the accuracy and safety of nutritional intervention.

[0030] 3. This invention, through the deep integration of knowledge graphs and game theory logic, can identify strategic nutrient combinations that not only promote the colonization of beneficial bacteria but also effectively inhibit competitors, producing a precise regulatory effect far exceeding that of traditional recommendation logic. This provides a highly intelligent decision support tool for personalized precision nutrition and chronic disease prevention.

[0031] 4. The architecture design of this invention fully considers the dynamic nature of biofeedback, has self-optimization and iteration capabilities, and can continuously improve the accuracy of prediction as user data accumulates, thus having extremely high clinical application value and social benefits. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall technical solution architecture according to the present invention; Figure 2 This is a schematic diagram of the core principle framework according to the present invention; Figure 3 This is a logical flowchart of the construction of heterogeneous semantic knowledge graph and multidimensional data feature extraction in the microecological field according to the present invention; Figure 4 This is a schematic diagram of the multi-level interaction relationships and data flow generated by the system steady-state identification and strategic nutrient recommendation based on Nash equilibrium point calculation according to the present invention. Figure 5 This is a schematic diagram illustrating the principle of the profit function weight adjustment and system parameter self-iteration based on the biological feedback closed loop in this invention. Detailed Implementation

[0033] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0034] The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system includes a microecological knowledge graph construction device, a user gut sample data collection device, a microbiota game relationship modeling device, a dynamic competitive evolution reasoning device, a Nash equilibrium point calculation device, a colonization probability prediction device, a strategic nutrient recommendation output device, a feedback closed-loop optimization device, and an environmental parameter correction device.

[0035] The aforementioned microbial ecosystem knowledge graph construction device is used to construct a heterogeneous semantic network containing microbial species nodes, metabolite nodes, nutrient nodes, and functional gene nodes by integrating biomedical literature, microbial metabolism databases, and clinical nutrition research data. The device integrates a natural language processing engine configured to extract microbial entities from massive amounts of unstructured text data using named entity recognition algorithms and to identify logical relationships between different entities using relation extraction algorithms.

[0036] During the construction process, the microbial ecosystem knowledge graph construction device defines a multi-level category system using a preset ontology model. Among them, the microbial species node includes microbial entities of different classification levels such as genus, species and strain, the metabolite node includes short-chain fatty acids, neurotransmitters, bile acids and various intermediate metabolites, and the nutrient node covers monosaccharides, polysaccharides, proteins, amino acids, lipids and trace elements.

[0037] The microbial ecosystem knowledge graph construction device identifies the interaction relationships between different bacterial communities through a preset entity extraction logic. These interaction relationships include mutualistic symbiosis, competitive inhibition, predation, and neutral metabolic associations. The device further refines these associations into communication associations mediated by signal molecules, energy flow associations based on metabolite coupling, and physical competition associations based on steric hindrance.

[0038] For each type of interaction, the microecological knowledge graph construction device assigns specific logical reasoning rules. For example, in the metabolic product coupling association, the rule is defined as follows: if the metabolic product of the first bacterial species is an essential growth substrate for the second bacterial species, then it is determined that there is a one-way mutual benefit relationship between the first bacterial species and the second bacterial species, and a corresponding relationship strength score is assigned.

[0039] The microbial ecosystem knowledge graph construction device further assigns biological attributes to each microbial node. These biological attributes include growth rate constant, substrate affinity coefficient, and survival threshold under specific conditions. These attributes are stored in the node's feature attribute vector as basic parameters for subsequent calculation of evolutionary game payoffs.

[0040] The user gut sample data acquisition device is used to acquire raw gut microbiome sequencing data and clinical physiological indicators of the target user. The user gut sample data acquisition device includes a metagenomics data interface, a metabolomics data interface, and a clinical indicator input unit.

[0041] The metagenomics data interface is configured to receive standardized files containing 16S ribosomal DNA sequencing data or whole-metagenomic shotgun sequencing data, and to invoke preset bioinformatics workflows for quality control, sequence assembly, and species annotation. The metabolomics data interface is used to receive metabolite abundance matrices generated by mass spectrometry or nuclear magnetic resonance.

[0042] The user intestinal sample data acquisition device standardizes the raw data, including logarithmic transformation and centering of abundance data, to identify the distribution of dominant bacteria, abundance of potential pathogens, and bacterial diversity indicators in the user's gut. The device then transforms the processed data into an initial state feature vector recognizable by the system. This initial state feature vector describes the initial resource allocation and bacterial community topology of the user's gut microbiota, serving as the input benchmark for subsequent game theory deduction.

[0043] The aforementioned microbial community game-theoretic modeling device is used to map beneficial, harmful, and neutral bacteria in the gut to independent players in a game theory model based on the semantic network generated by the microbial ecosystem knowledge graph construction device. The device integrates a player attribute definition module, configured to assign individual characteristics such as resource utilization ability, metabolic flexibility, and environmental tolerance to each microbial player based on biological attributes in the knowledge graph. When simulating player behavior, the device introduces the bounded rationality assumption, considering the predetermined delays and non-optimal response characteristics of microorganisms in complex biochemical environments, enabling the game model to reflect the adaptive adjustment process of microorganisms during resource competition.

[0044] The microbial community game theory modeling device maps selectable nutrient combinations to player-selectable resource acquisition strategies, and transforms the colonization abundance, metabolite production, and immune contribution of the microbial community under specific nutrient intervention into corresponding game payoff functions. When defining the payoff function, the device uses the microbial community's substrate utilization rate of specific nutrients, resistance to harmful bacterial secretions, and adhesion ability to the intestinal mucosa as positive payoff modifiers, while using environmental stress coefficients and metabolic waste inhibition effects as negative payoff modifiers.

[0045] The calculation logic for the payoff function is configured as follows: ; Represented as instantaneous gain, this is the instantaneous gain of the bacterial community under a specific strategy combination; Expressed as substrate utilization rate; The weighting coefficient corresponding to the weighting coefficient of substrate utilization; This is represented as a resistance score; This is expressed as the weighting coefficient of the resistance score, and the weighted sum of the resistance scores; This is expressed as adhesive ability; This is expressed as a weighting coefficient of adhesion ability, or a weighted sum of adhesion abilities; This is expressed as the negative effect value caused by environmental stress; Expressed as the concentration of metabolic waste; It is expressed as the inhibition coefficient of metabolic waste.

[0046] The dynamic competitive evolutionary inference device is used to simulate the dynamic confrontation process between players under different nutrient delivery strategies. This device breaks away from traditional static weighted inference logic, introducing an evolutionary game theory algorithm to describe the strategy iteration path of a microbial community in a limited space and resource environment.

[0047] The dynamic competitive evolutionary reasoning device integrates a dynamic evolutionary simulation engine, which is configured to simulate the topological changes of different bacterial community nodes in the process of resource grabbing in real time by constructing a dynamic competitive network. This topological change reflects the dynamic migration of the competitive intensity between bacterial communities. For example, when a certain nutrient is consumed in large quantities by a certain bacterial community, the connection weight of other bacterial community nodes that have overlapping resources with it will decrease.

[0048] The dynamic competitive evolutionary inference device is equipped with replicated dynamic equation logic to describe the rate of change in the proportion of a specific bacterial population during competition. The specific execution logic of the replicated dynamic equation logic is configured as follows: first, calculate the expected return of a specific bacterial population player under the current strategy; then, obtain the average return of all bacterial population players in the system. Next, calculate the difference between the expected return and the average return. If the difference is positive, it indicates that the strategy of this bacterial population is better than the average level, and its proportion in the intestinal system will increase with time step; if the difference is negative, its proportion will decrease accordingly. The dynamic competitive evolutionary inference device utilizes high-dimensional matrix operation logic to process pairwise game relationships between large-scale bacterial populations, iteratively calculates the population proportion distribution within each discrete time step, and summarizes them into an evolutionary trend map at the population level.

[0049] The Nash equilibrium point calculation device is used to perform convergence analysis on the evolutionary trajectory generated by the dynamic competitive evolutionary reasoning device to find the system's stable point in a multi-party game environment. The Nash equilibrium point calculation device is internally equipped with a nonlinear stability analysis module, which identifies, through iterative calculation, the final stable state distribution of the microbial community system under specific nutrient intervention conditions. A stable state refers to a distribution where no single microbial community can obtain higher returns by changing its resource utilization strategy; that is, the system has reached a game equilibrium.

[0050] The Nash equilibrium point calculation device employs multi-objective optimization logic, simultaneously considering maximizing the colonization abundance of beneficial bacteria, maximizing the growth inhibition rate of harmful bacteria, and minimizing the host's metabolic load when searching for an equilibrium point. The Nash equilibrium point calculation device can identify multiple equilibrium states and, based on preset optimal health standards, select the equilibrium point from multiple potential stable states that offers the highest degree of improvement to the host's physiological indicators.

[0051] The Nash equilibrium point calculation device is specifically equipped with risk identification logic to determine whether, upon reaching an equilibrium state, harmful bacteria have gained more than a preset threshold by competing for nutrient resources. For example, if the system predicts that the supplementation of a certain prebiotic will lead to an explosive growth of some facultative pathogens using alternative metabolic pathways, causing the final equilibrium ratio of pathogens to exceed the safety warning value, then the risk identification logic marks this state as a nutrient backlash risk state.

[0052] The colonization probability prediction device is used to quantify the colonization success rate of the target beneficial bacteria within the prediction period based on the output of the Nash equilibrium point calculation device. The colonization probability prediction device is equipped with time-dimensional simulation logic to predict the persistence of bacterial colonization and its robustness against external interference at different time spans.

[0053] The colonization probability prediction device assesses the degree of improvement of the current nutrient intervention strategy on the gut microbiota balance by comparing the predicted equilibrium state with a preset healthy gut standard model. The prediction logic is configured to compare the abundance of target beneficial bacteria in the equilibrium state with a preset successful colonization threshold. If it is higher than the threshold, a colonization success probability score is calculated based on the difference, and the success probability score is corrected by combining the system diversity index to output the final percentage probability.

[0054] The strategic nutrient recommendation output device is used to generate and output a personalized nutrient combination plan for the target user when the colonization probability prediction device determines that the colonization success rate is higher than a preset probability threshold, and the Nash equilibrium point calculation device determines that the system converges to an equilibrium state where beneficial bacteria dominate. The plan generated by the strategic nutrient recommendation output device includes recommended prebiotic types, specific amino acid ratios, trace element content, and suggested supplementation cycles.

[0055] Before outputting a plan, the strategic nutrient recommendation output device automatically performs a safety screening. If the prediction results show that the addition of a certain nutrient will cause harmful bacteria to grow explosively through a specific metabolic pathway, or cause the production of harmful metabolites such as ammonia and hydrogen sulfide to exceed the safety standard, the output of the plan will be blocked immediately, and the dynamic competitive evolutionary reasoning device will be called back to find an alternative strategy.

[0056] The feedback closed-loop optimization device is used to acquire secondary intestinal testing data after the user executes the recommended plan and compare it with the previous prediction results. The device automatically adjusts the weights of the payoff function in the microbial community game modeling device and the evolutionary parameters in the dynamic competitive evolutionary inference device by calculating the prediction deviation—the Euclidean distance between the actual microbial abundance change and the predicted equilibrium state. For example, if the actual observed growth rate of a certain microbial community is lower than the predicted value, the substrate utilization weight of that community in the payoff function is reduced, enabling the system model to self-iterate.

[0057] The environmental parameter correction device is used to personalize the game payoff function based on the user's age, gender, region, and basal metabolic index. The device includes a personalized deviation matrix configured to store physiological environmental constants under different population characteristics. For example, for elderly users, it automatically lowers the baseline value of the intestinal mucosal adhesion coefficient to ensure the game model reflects the differences in microbial competition under different host backgrounds.

[0058] Example 2: Based on Example 1, this example provides a knowledge graph-based system for predicting gut microbiota colonization effects and recommending nutrients, based on an edge computing and cloud collaborative architecture. The system aims to improve the processing efficiency of large-scale concurrent users and ensure the real-time performance of data transmission.

[0059] The system includes an intelligent data acquisition terminal deployed on the user side and a deep inference center deployed on a cloud server. The intelligent data acquisition terminal integrates the hardware interface of the user's intestinal sample data acquisition device and is equipped with a lightweight data preprocessing unit. This data preprocessing unit is configured to perform basic filtering and compression on the acquired metagenomic sequencing raw signals, and use hash mapping technology to convert complex gene sequences into fixed-length compact fingerprints to reduce the amount of data transmitted to the cloud.

[0060] The cloud server is equipped with a high-performance computing cluster to run the microbial ecosystem knowledge graph construction device, the microbial community game relationship modeling device, the dynamic competitive evolution reasoning device, and the Nash equilibrium point calculation device. In the cloud architecture, the microbial ecosystem knowledge graph construction device uses a distributed graph database for storage, supporting parallel queries of hundreds of millions of nodes and edge relationships. The dynamic competitive evolution reasoning device utilizes a graphics processing unit array to perform large-scale matrix evolution operations, capable of simulating the microbial community game process under thousands of nutrient combinations in parallel.

[0061] The described microbial community game modeling device is configured to support multi-player concurrent game models in a cloud environment. For each individual user's gut microbiota, the cloud server assigns an independent virtual evolution container. Within the virtual evolution container, the system subdivides the gut microbiota into different functional groups, each group serving as a game subject. The calculation of the payoff function not only relies on the biological attributes described in Embodiment 1 but also incorporates real-time updated global microbial epidemiological data from the cloud as a global prior distribution.

[0062] When executed in the cloud, the dynamic competitive evolutionary inference device employs evolutionary logic based on asynchronous random game theory. This logic is configured to: in each evolutionary step, randomly select a subset of microbial community nodes for policy updates, simulating random encounters and resource competition among microorganisms in the local environment of the gut. By introducing random noise, the dynamic competitive evolutionary inference device can more accurately simulate the volatility of the real gut environment and identify colonization strategies that remain robust under environmental disturbances.

[0063] The Nash equilibrium point calculation device employs a distributed gradient search algorithm in the cloud to locate the system's evolutionary stability strategy. When multiple equilibrium points exist, the calculation device is equipped with global optimization logic, which calculates the host health energy score corresponding to each equilibrium point and selects the globally optimal solution. The host health energy score is calculated based on the semantic weights in the knowledge graph relating microbial metabolites to human physiological health; for example, equilibrium states producing more short-chain fatty acids such as butyrate will be assigned higher energy scores.

[0064] The feedback loop optimization device in this embodiment employs an incremental learning-based update mechanism. When a large amount of secondary detection data from users is collected in the cloud, the feedback loop optimization device no longer retrains the entire game model; instead, it uses an incremental learning algorithm to fine-tune the existing payoff function parameters. This mechanism enables the system to continuously evolve as the sample size increases, identifying unique microbial community game patterns within specific populations.

[0065] After generating the recommended nutrient scheme in the cloud, the device sends the results to the user's smart data collection terminal via an encrypted transmission protocol. The terminal is equipped with an interactive display unit that transforms complex game-theoretic trends into intuitive visual charts, demonstrating how the recommended scheme guides beneficial bacteria to win in the competition and highlighting potential risks of harmful bacteria outbreaks, thereby increasing user trust and compliance with the recommended scheme.

[0066] Example 3: Based on the above examples, this example describes in detail a gut microbiota colonization effect prediction and nutrient recommendation system based on federated learning for multi-center collaborative scenarios, aiming to resolve the contradiction between data privacy protection and knowledge graph sharing among different medical institutions.

[0067] The system includes multiple local computing nodes distributed in different geographical locations and a central coordination server. Each local computing node is deployed in a different clinical research institution or medical center, and each contains all the devices described in Example 1, but the data it processes is limited to user samples collected by that institution.

[0068] The microecological knowledge graph construction device is configured in a two-layer structure on the local computing node: a basic general layer and a local feature layer. The basic general layer obtains publicly available biomedical common sense from the central coordination server, while the local feature layer utilizes institution-specific clinical observation data for enhancement.

[0069] When the microbial community game relationship modeling device is executed on a local node, it exchanges parameters with the game models of other nodes through a federated learning protocol issued by the central coordination server. Specifically, the local nodes do not directly exchange the user's original microbial community data, but rather exchange the gradient parameters of the game payoff function. The central coordination server is configured with a secure aggregation algorithm to perform a weighted average of the gradients of the payoff functions from each local node.

[0070] The dynamic competitive evolutionary inference device, within a federated learning framework, is configured to perform local evolutionary inference. Each local node simulates the dynamic competition process of the microbial community on its local dataset based on the aggregated globally optimal parameters. Through multiple rounds of local inference-parameter uploading-global aggregation-parameter distribution loops, the system can ultimately train a microbial community evolution prediction model with extremely high generalization ability while protecting data privacy.

[0071] In addition to considering the equilibrium conditions described in Example 1, the Nash equilibrium point calculation device introduces a group consensus constraint in the local node. This group consensus constraint logic is configured to: when searching for a local equilibrium point, calculate the deviation between the local equilibrium state and the global average equilibrium state, and add it as a penalty term to the optimization objective. This ensures that the nutrient schemes generated by the recommendation system are not only effective on the current local sample but also conform to the general principles of microbial ecology.

[0072] After generating results at the local node, the strategic nutrient recommendation output device invokes the cross-validation logic of the central coordination server. This cross-validation logic is configured to use anonymized validation sets generated by other nodes to robustly evaluate the current node's recommendation scheme. Only schemes that pass cross-validation are ultimately output to the user.

[0073] The feedback loop optimization device in this embodiment also operates in federated learning mode. When a local institution collects user feedback data, its calculated model update increment is sent to the central coordination server via homomorphic encryption. The central server completes the parameter update in encrypted mode, ensuring that even in the event of a system intrusion, the privacy data of individual users' microecological systems will not be leaked. This architecture is particularly suitable for multi-center clinical trials, enabling rapid integration of dispersed microecological intervention data and accelerating the development of precision nutrition strategies.

[0074] Example 4: This example describes in detail the data flow mechanism, hardware collaboration logic, and complex logical decision branches between the various devices in the system, in order to further demonstrate the overall operating principle of the system.

[0075] During the system startup phase, the microbial ecosystem knowledge graph construction device first retrieves raw triplet data from a distributed biomedical database. The triplet is defined as: subject microbial species, predicate relation, and guest metabolite or guest nutrient. The knowledge fusion unit within the device is configured to: identify and merge redundant entities from different data sources, calculate the semantic distance between entities using a cosine similarity algorithm, and determine if the similarity is higher than a preset threshold of 0.95, classifying them as the same entity. The processed structured knowledge is mapped into a high-dimensional sparse matrix and stored in the system's global knowledge base memory.

[0076] When a user submits a sample through the user gut sample data acquisition device, the system's instruction flow is activated. The metagenomic analysis module in the data acquisition device is configured to call an external high-performance computing core to perform sequence alignment tasks. During the alignment process, the module uses a Bloom filter to quickly exclude non-microbial sequences and employs a hidden Markov model to predict functional genes in the gut microbiota. The generated abundance distribution table is passed to a memory buffer, triggering the operation of the gut microbiota game relationship modeling device.

[0077] The microbial community game relationship modeling device reads microbial community abundance data from a memory buffer and dynamically generates a directed weighted game graph based on the interaction relationships in a global knowledge base. In the game graph, nodes represent different functional microbial communities, and the weights of edges represent the strength of their interactions. The modeling device then initializes a payoff matrix. Each cell of the payoff matrix is ​​configured to store the expected payoff value for a specific microbial community when adopting a specific resource strategy.

[0078] The computational logic includes the following decision branches: Decision Branch A: Detect whether the current bacterial community possesses genes for specific metabolic pathways that utilize the target nutrient. If so, the corresponding substrate utilization coefficient is set to a positive value; otherwise, it is set to zero. Decision Branch B: Detect whether there are inhibitory factors against this bacterial community in the environmental parameters. If so, the benefit reduction component is calculated based on the concentration of the inhibitory factor. Decision Branch C: Detect whether competing bacterial communities produce secondary metabolites that are harmful to this bacterial community. If so, the potential expected loss is calculated based on the initial abundance of competing bacterial communities.

[0079] The results of the above-mentioned decision branches are weighted and summed to form a complete initial payoff matrix, which is then transmitted to the dynamic competitive evolution reasoning device.

[0080] The dynamic competitive evolutionary inference device initiates time-step iteration. Within each iteration cycle, the system performs the following operations: Operation 1: Calculate the average benefit of each bacterial community under the current population distribution.

[0081] Operation 2: Invoke the replication dynamic equation logic to calculate the incremental percentage of each bacterial community at the next time step. This increment is equal to the percentage of the bacterial community multiplied by the difference between its instantaneous gain and average gain, and then divided by the total damping coefficient of the system, to simulate the physical constraint of intestinal volume on bacterial community expansion.

[0082] Operation 3: Update the topology of the dynamic competition network. If the proportion of a certain bacterial community is less than 0.0001, it is marked as temporarily dormant and its computational priority is reduced in subsequent iterations.

[0083] Step 4: Detect the change in the total kinetic energy of the system. If the sum of the rates of change of the proportion of all bacterial communities is less than the preset value of 0.0005 within 100 consecutive steps, the system is determined to have reached stability and the iteration is stopped.

[0084] The state vector after evolution terminates is passed to the Nash equilibrium point calculation device. This device first performs a convergence validity check. If it finds the system is in a limit cycle oscillation rather than fixed-point convergence, it automatically triggers a perturbation restart logic, fine-tunes the initial nutrient delivery strategy, and restarts the evolution. If convergence to a fixed point is confirmed, the system proceeds to the risk screening stage.

[0085] In the risk screening process, the Nash equilibrium point calculation device is equipped with sensitivity analysis logic. This logic observes the direction of the equilibrium point's position movement by making small perturbations to the nutrient concentration. If a small increase in concentration is found to cause a non-linear, jump-like increase in the abundance of pathogenic bacteria, the system determines that the equilibrium point has a risk of catastrophic reversal, and immediately blocks the recommended path.

[0086] The colonization probability prediction device takes over the data stream. This device uses a pre-trained deep convolutional neural network to extract features from the evolutionary trajectory. The neural network is configured to take an abundance time series graph during the evolutionary process as input and a colonization success rate as output. By capturing the dynamic stability features during the evolutionary process, the device can output a prediction probability score with extremely high confidence.

[0087] The strategic nutrient recommendation output device generates a final personalized report based on the predicted score and balanced health score. The report is pushed to the user through the system's output interface. All parameters of the recommendation scheme are saved to the feedback log database.

[0088] When the feedback loop optimization device receives a new round of detection data from the user, the system performs a prediction-measurement consistency check. If the deviation exceeds a preset range, the system automatically initiates a self-healing update logic. This self-healing update logic is configured to use a gradient descent algorithm to search within the parameter vector space of the reward function for the parameter combination that best approximates the previous prediction results to the actual observations. The updated parameters will overwrite the original default parameters and be stored in the personalized configuration file of the environmental parameter correction device, completing the closed-loop evolution of the system.

[0089] Example 5: This example further illustrates the hardware deployment scheme and specific functional configuration of the system in different application scenarios.

[0090] Scenario 1: Management of long-term chronic inflammation. The environmental parameter correction device is configured to strengthen the coupling relationship between inflammatory factors and the benefits of gut microbiota interaction. The system accesses the user's blood biochemical indicators via an interface. The gut microbiota interaction modeling device maps high concentrations of inflammatory factors to an additional survival pressure coefficient for beneficial bacteria. The recommended output device will prioritize prebiotic combinations with anti-inflammatory properties, and Nash equilibrium calculations will ensure that these prebiotics are not preempted by pro-inflammatory bacteria.

[0091] Scenario 2: Maintaining Gut Homeostasis in Athletes. In this scenario, the system incorporates a sports metabolism module. This module, connected to the environmental parameter correction device, acquires the user's exercise intensity, perspiration volume, and lactic acid metabolism data. During the simulation, the dynamic competitive evolutionary inference device introduces periodic physical perturbation factors to simulate the instantaneous impact of changes in intestinal blood perfusion caused by high-intensity exercise on the gut microbiota distribution. The system-generated nutrient recommendation scheme focuses on rapidly repairing the damaged intestinal barrier and promoting a balance in the interplay between lactogenic and lactic acid-consuming bacteria.

[0092] Scenario 3: Early Intestinal Colonization in Infants and Young Children. In this scenario, the microbiome knowledge graph construction device loads a specialized developmental microbiome sub-graph. In this developmental microbiome sub-graph, predation relationships and steric hindrance competition among microbiota are given higher weights. When assessing the success rate, the colonization probability prediction device extends the time dimension to a longer prediction period to evaluate the profound impact of early nutritional intervention on lifelong microbiome patterns.

[0093] At the hardware implementation level, the aforementioned devices can be integrated into a single dedicated server or deployed in a distributed manner within a private cloud. The dynamic competitive evolutionary reasoning device and Nash equilibrium point calculation device, which involve large-scale matrix operations, are recommended to be configured with graphics processing units (GPUs) with general-purpose computing capabilities to utilize their thousands of cores for parallel processing of the game's dynamic equations. The knowledge graph construction device is recommended to use a server equipped with large-capacity non-volatile memory to support frequent graph updates and high-concurrency semantic retrieval operations. Data exchange between the devices employs a binary format based on an efficient protocol buffer to minimize communication latency.

[0094] The system described in this embodiment, by deeply integrating the dynamic thinking of evolutionary game theory into the static structure of the knowledge graph, not only solves the problem of knowing what but not why in the regulation of the micro-ecosystem, but also provides an unprecedented predictive dimension for precision nutrition through the quantitative deduction of the adversarial competition logic, avoiding the risk of ecological imbalance caused by blind supplementation, and demonstrating extremely high clinical applicability and technological advancement.

[0095] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0096] In this document, unless otherwise expressly specified and limited, the terms installation, connection, linking, fixing, etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0097] In this embodiment of the invention, unless otherwise explicitly specified and limited, the first feature above or below the second feature may be in direct contact with the first feature, or indirect contact via an intermediate medium. Furthermore, "above," "over," and "on top" of the first feature may mean the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "beneath" of the first feature may mean the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0098] In the description of this specification, references to the terms "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with this embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. The specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

Claims

1. A knowledge graph-based system for predicting gut microbiota colonization and recommending nutrients, characterized in that, include: A knowledge graph construction device in the field of microbial ecology is used to construct a heterogeneous semantic network containing microbial species nodes, metabolite nodes, nutrient nodes and functional gene nodes by integrating biomedical literature, microbial metabolism databases and clinical nutrition research data. The user gut sample data acquisition device is used to acquire raw gut microbial sequencing data and clinical physiological indicators of the target user and convert them into an initial state feature vector. A microbial community game relationship modeling device is used to map beneficial bacteria, harmful bacteria and neutral bacteria in the gut as game players based on the heterogeneous semantic network, and to map the available nutrient combinations as resource acquisition strategies to construct a game payoff function. The dynamic competitive evolutionary reasoning device is used to simulate the dynamic confrontation process between players under different nutrient delivery strategies, and to deduce the evolutionary trend map of the gut ecosystem through evolutionary game algorithm; The Nash equilibrium point calculation device is used to perform convergence analysis on the evolution trajectory, identify the stable point of the system, and is equipped with risk identification logic to determine whether the benefit growth of harmful bacteria in the equilibrium state exceeds a preset threshold. A colonization probability prediction device is used to quantify the colonization success rate of target beneficial bacteria within the prediction period based on the output results of the system's stable point. The strategic nutrient recommendation output device is used to generate a personalized nutrient combination plan for the user when the colonization success rate meets a preset probability threshold and the system converges to an equilibrium state where beneficial bacteria dominate.

2. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 1, characterized in that: The micro-ecological knowledge graph construction device integrates a natural language processing engine, which is configured to extract micro-ecological entities from unstructured text data using a named entity recognition algorithm and to identify logical relationships between different entities using a relation extraction algorithm. The microbial ecosystem knowledge graph construction device uses a preset ontology model to define a multi-level category system, in which the microbial species node includes genus-level, species-level and strain-level microbial entities, the metabolite node includes short-chain fatty acids, neurotransmitters, bile acids and intermediate metabolites, and the nutrient node covers monosaccharides, polysaccharides, proteins, amino acids, lipids and trace elements. The microbial ecosystem knowledge graph construction device identifies the interaction relationships between different bacterial communities through a preset entity extraction logic. These interaction relationships include mutualistic relationships, competitive inhibition relationships, predatory relationships, and neutral metabolic associations. The microecological knowledge graph construction device further refines the associations into communication associations mediated by signal molecules, energy flow associations based on metabolite coupling, and physical competition associations based on steric hindrance, and assigns logical reasoning rules to each type of interaction. The microbial ecosystem knowledge graph construction device further assigns biological attributes to each microbial node. These biological attributes include growth rate constant, substrate affinity coefficient, and survival threshold in the environment. The biological attributes are stored in the node's feature attribute vector.

3. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 2, characterized in that: The user intestinal sample data acquisition device includes a metagenomics data interface, a metabolomics data interface, and a clinical indicator input unit. The metagenomic data interface is configured to receive standardized files containing 16-bit ribosomal DNA sequencing data or whole metagenomic shotgun sequencing data, and to invoke preset bioinformatics workflows for quality control, sequence assembly, and species annotation. The metabolomics data interface is used to receive metabolite abundance matrices generated by mass spectrometry or nuclear magnetic resonance. The user intestinal sample data acquisition device performs standardization processing on the raw data. The standardization processing includes logarithmic transformation and centering of the abundance data to identify the distribution of dominant bacteria, abundance of potential pathogens, and bacterial diversity indicators in the user's intestines. The user intestinal sample data acquisition device converts the processed data into an initial state feature vector. The initial state feature vector is used to describe the initial resource allocation state and microbial community structure topology of the user's intestinal micro-ecosystem, and serves as the input benchmark for subsequent game theory deduction.

4. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 3, characterized in that: The microbial community game relationship modeling device integrates a player attribute definition module, which is configured to assign resource utilization ability, metabolic flexibility and environmental tolerance characteristics to each microbial community player based on the biological attributes in the heterogeneous semantic network. When simulating player behavior, the microbial community game relationship modeling device introduces the bounded rationality assumption and simulates the adaptive adjustment process of microorganisms in the process of resource grabbing by setting predetermined reaction delay parameters and non-optimal response characteristic parameters. When defining the payoff function, the microbial community game relationship modeling device uses the substrate utilization rate of nutrients, resistance to harmful bacterial secretions, and adhesion ability of the microbial community to the intestinal mucosa as positive correction factors for payoff, and uses the environmental pressure coefficient and metabolic waste inhibition effect as negative correction factors for payoff. The calculation logic for the payoff function is configured as follows: Multiply the substrate utilization rate by the corresponding first weighting coefficient, add the product of the resistance score and the second weighting coefficient, add the product of the adhesion ability and the third weighting coefficient, subtract the negative effect value caused by environmental stress, and subtract the product of metabolic waste concentration and inhibition coefficient to finally output the instantaneous benefit of the microbial community under a specific strategy combination.

5. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 4, characterized in that: The dynamic competitive evolutionary reasoning device integrates a dynamic evolutionary simulation engine. The engine is configured to simulate the topological changes of different bacterial community nodes in the process of resource grabbing by constructing a dynamic competitive network. By calculating the degree of decay of the connection weight of other bacterial community nodes that have overlapping resources when a certain nutrient is consumed by a certain bacterial community, the dynamic migration of the competitive intensity between bacterial communities is reflected. The dynamic competitive evolution reasoning device is equipped with a copy dynamic equation logic. The specific execution steps of the copy dynamic equation logic include: First, calculating the expected returns of the swarm players under the current strategy. The second step is to obtain the average earnings of all players in the system's microbiome. The third step is to calculate the difference between the expected return and the average return. If the difference is positive, it is determined that the strategy of the bacterial community is better than the average level, and the increase in the proportion of the bacterial community in the system is equal to the product of the difference, the current proportion and the reciprocal of the total damping coefficient of the system. If the difference is negative, its proportion is determined to decrease accordingly; The dynamic competitive evolutionary reasoning device uses high-dimensional matrix operation logic to process the pairwise game relationship between bacterial communities, calculates the population proportion distribution within each discrete time step through iterative calculation, and summarizes them into an evolutionary trend map at the population level.

6. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 5, characterized in that: The Nash equilibrium point calculation device is equipped with a nonlinear stability analysis module, which identifies the final stable state distribution of the microbial community system under nutrient intervention conditions through iterative calculation. The stable state is defined as a game equilibrium state in which no microbial community can obtain higher returns by changing its resource utilization strategy. The Nash equilibrium point calculation device adopts multi-objective optimization logic, which simultaneously considers maximizing the colonization abundance of beneficial bacteria, maximizing the growth inhibition rate of harmful bacteria, and minimizing the host metabolic load when searching for the equilibrium point. The Nash equilibrium point calculation device is equipped with sensitivity analysis logic. By perturbing the nutrient concentration, it observes the direction of the equilibrium point's position movement. If the perturbation is detected to cause a non-linear increase in the equilibrium abundance of pathogenic bacteria, it is determined that the equilibrium point has a risk of nutrient backlash. The computing device can identify multiple equilibrium states and, based on a preset optimal health standard, select the equilibrium point that provides the highest degree of improvement to the host's physiological indicators from multiple potential stable states. If the system is found to be in a limit cycle oscillation rather than a fixed-point convergence, the perturbation restart logic is automatically triggered, and the evolution is restarted by fine-tuning the initial nutrient delivery strategy.

7. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 6, characterized in that: The colonization probability prediction device is equipped with time-dimensional simulation logic to predict the persistence of microbial colonization and its robustness against external interference over different time spans. The colonization probability prediction device assesses the degree of improvement of the gut microbiota balance by comparing the predicted equilibrium state with a preset healthy gut standard model. The colonization probability prediction device integrates a pre-trained deep convolutional neural network. The neural network is configured to take the abundance time series diagram in the evolution process as input, capture the resilience characteristics and damping ratio characteristics in the evolution process, and quantify and output the colonization success rate score. The specific prediction logic is configured as follows: compare the abundance of the target beneficial bacteria in the equilibrium state with the preset successful colonization threshold. If it is higher than the successful colonization threshold, calculate the basic score based on the difference between the abundance of the target beneficial bacteria and the threshold, and then correct the basic score in combination with the system diversity index, and finally output the percentage probability.

8. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 7, characterized in that: The strategy nutrient recommendation output device generates a plan that includes the recommended types of prebiotics to be supplemented, the amino acid ratio, the content of trace elements, and the suggested supplementation cycle. Before outputting a plan, the strategic nutrient recommendation output device will automatically perform a safety screening. If the prediction results show that the addition of a certain nutrient will cause harmful bacteria to grow explosively through metabolic pathways, or cause the production of harmful metabolites to exceed the preset safety standard, the output of the plan will be blocked immediately, and the dynamic competitive evolution reasoning device will be called back to find an alternative strategy. The strategic nutrient recommendation output device pushes the generated personalized report to the mobile application interface or medical terminal interface through the system's output interface, while saving all parameters of the recommendation plan to the feedback log database.

9. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 8, characterized in that, Also includes: The feedback closed-loop optimization device is used to acquire the user's secondary intestinal test data after implementing the recommended plan, and compare and analyze it with the previous prediction results; The feedback closed-loop optimization device calculates the prediction deviation, i.e., the Euclidean distance between the actual changes in bacterial abundance and the predicted equilibrium state, and uses the gradient descent algorithm to search in the parameter vector space of the payoff function. It automatically adjusts the payoff function weights in the bacterial community game relationship modeling device and the evolution parameters in the dynamic competitive evolution inference device to achieve self-iteration of the system model. An environmental parameter correction device is used to perform personalized parameter calibration of the game payoff function based on the user's age, gender, region, and basal metabolic index. The environmental parameter correction device is equipped with a personalized deviation matrix, which is configured to store physiological and environmental constants under different population characteristics. When the user is a pre-defined elderly population, the environmental parameter correction device automatically lowers the baseline value of the intestinal mucosal adhesion coefficient to ensure that the game model reflects the differences in microbial competition under different host backgrounds.

10. The knowledge graph-based gut microbiota colonization effect prediction and nutrient recommendation system according to claim 9, characterized in that: The system adopts a distributed collaborative architecture, including intelligent data collection terminals deployed on the user side and a deep inference center deployed on a cloud server; The intelligent acquisition terminal is equipped with a lightweight data preprocessing unit, which is used to filter and compress the raw signal, and to use hash mapping technology to convert the gene sequence into a fixed-length compact fingerprint. The cloud server is equipped with a high-performance computing cluster for running the micro-ecological knowledge graph construction device, the microbial community game relationship modeling device, the dynamic competitive evolution reasoning device, and the Nash equilibrium point calculation device. The feedback closed-loop optimization device adopts an update mechanism based on incremental learning. After the secondary detection data of multiple users are collected in the cloud, the incremental learning algorithm is used to fine-tune the existing profit function parameters. The system is further configured with a federated learning protocol, which uses a central coordination server to securely aggregate the gradients of the profit function from each local node without exchanging the user's original data, and uses homomorphic encryption technology to ensure privacy and security during the parameter update process.

Images