A pig immune function prediction system and method based on artificial intelligence

Abstract

The present invention provides an artificial intelligence-based system and method for predicting the immune function of pigs, comprising a detection device for detecting intestinal flora detection data corresponding to multiple omics and detection result data of multiple metabolic markers of pigs under stress; an edge optimization decision device for performing feature extraction based on the detection result data to obtain metabolic characteristic data; a prediction and display device for simulating the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules of pigs under stress based on the intestinal flora detection data to obtain metabolic immune interaction data, storing the metabolic immune interaction data in a metabolic immune database, determining the dose-effect relationship between functional amino acids and immune phenotypes based on the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, obtaining the metabolic spectrum characteristics of pigs under stress, and predicting, generating and displaying amino acid ratios or dietary intervention strategies based on the metabolic spectrum characteristics.

Description

Technical Field

[0001] The present disclosure relates to the field of animal husbandry technology, and in particular to a pig immune function prediction system and method based on artificial intelligence. Background Art

[0002] With the rapid advancement of intensive, large-scale and intelligent breeding models, pig herds are facing pathogen infection stress, early weaning stress, etc., which lead to imbalance of the body's immune homeostasis and significant decline in disease resistance, thereby causing disease outbreaks and abnormal deaths.

[0003] Amino acid metabolism disorders are a core factor affecting the immune function of pigs, especially under weaning stress or infection conditions. Intestinal amino acid metabolic remodeling directly regulates immune cells, thereby affecting intestinal barrier function and systemic immune response. However, traditional nutritional regulation methods are mostly based on empirical formulas, lacking an analysis of the dynamic correlation mechanism between amino acid metabolism and immune function, resulting in insufficient intervention precision. At the same time, traditional methods rely on single metabolomic or immunological data, making it difficult to capture the dynamic network of metabolic-immune interactions; differences in amino acid requirements at different growth stages (such as weaning and fattening) and dietary patterns (such as low-protein diets) have not been quantified; and detection equipment relies on laboratory-level analysis, which cannot achieve rapid response in a farm environment. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a pig immune function prediction system based on artificial intelligence, the system comprising:

[0005] A detection device for detecting intestinal flora detection data corresponding to multiple omics and detection result data of multiple metabolic markers of the pig under stress;

[0006] An edge optimization decision-making device, configured to extract features based on the detection result data to obtain metabolic feature data;

[0007] A prediction and display device is used to simulate the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules of pigs under stress conditions based on a dynamic metabolic network according to the intestinal flora detection data, obtain metabolic immune interaction data, and store the metabolic immune interaction data in a metabolic immune database, and determine the dose-effect relationship between functional amino acids and immune phenotypes based on metabolic flow redirection according to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, so as to obtain the metabolic spectrum characteristics of the pigs under stress conditions. The metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios. According to the metabolic spectrum characteristics, amino acid ratios or dietary intervention strategies are predicted and generated.

[0008] In one possible implementation, the prediction and display device is specifically used to:

[0009] Based on the composition and abundance of the intestinal flora in the intestinal flora detection data and the metabolites corresponding to the intestinal flora, the nodes in the dynamic metabolic network are deleted to obtain a target metabolic network, wherein the dynamic metabolic network is constructed based on the full amount of metabolites and metabolic enzymes and includes metabolite nodes, metabolic enzyme nodes, and reaction edges, and the reaction edges are used to represent the conversion relationship between metabolites;

[0010] According to the regulation of the activity of the key metabolic enzyme by the metabolites corresponding to the intestinal flora, the key metabolic enzyme is added from the reaction edge through different metabolic enzyme nodes in the target metabolic network, and the expression of the immune effector molecule by the activity of the key metabolic enzyme of the pig under stress is simulated to obtain the correlation coefficient between the activity of the key metabolic enzyme and the expression of the immune effector molecule under different stress conditions;

[0011] A dynamic model based on flux balance analysis is used in combination with a two-level optimization algorithm to simulate changes in metabolic flux under stress conditions, and a mixed integer optimization method is used to extract the minimum equilibrium pathway from substrates to the metabolites to optimize the target metabolic network.

[0012] Based on the optimized target metabolic network, re-performing the step of adding key metabolic enzymes from the reaction edge through different metabolic enzyme nodes in the target metabolic network, simulating the effect of the activity of the key metabolic enzymes on the expression of the immune effector molecules under stress conditions, and obtaining correlation coefficients between the activity of the key metabolic enzymes and the expression of the immune effector molecules under different stress conditions, until the difference between the correlation coefficients obtained from two consecutive calculations is less than a preset coefficient threshold;

[0013] The metabolic immune interaction data is determined based on the addition amount of the key metabolic enzyme corresponding to the correlation coefficient obtained in the last calculation and the expression amount of the corresponding immune effector molecule.

[0014] In one possible implementation, the prediction and display device is specifically used to:

[0015] According to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, based on metabolic flux redirection, computer simulation technology is used to predict the dynamic changes of metabolic flux to obtain the concentration change data of the metabolite;

[0016] Predicting the metabolite accumulation data and functional amino acid consumption data of the pig under the stress state based on the metabolic flux and the metabolite concentration change data;

[0017] Based on the accumulation data of metabolites and the consumption data of functional amino acids under the stress state, the dose-effect relationship between the functional amino acids and the immune phenotype is determined to obtain the metabolic profile characteristics of the pigs under the stress state.

[0018] In one possible implementation, the detection device includes a near-infrared rapid detection module configured to:

[0019] determining infrared band data for spectrum acquisition based on characteristic absorption peaks of chemical bonds in the plurality of metabolic markers;

[0020] According to the infrared band data, spectrum acquisition is performed on the collected blood sample of the live pig to obtain a scanning spectrum;

[0021] performing characteristic absorption peak identification of the plurality of metabolic markers on the scanned spectrum to obtain a plurality of identification characteristic absorption peak data;

[0022] The multiple identification characteristic absorption peak data are substituted into a multivariate regression algorithm model, which is used to process the relationship between highly collinear spectral data and concentration to obtain detection result data of multiple metabolic markers.

[0023] In a possible implementation, the edge optimization decision-making device is specifically configured to:

[0024] determining target wavelength ranges for the multiple metabolic markers based on the vibration characteristics of the chemical bonds in the multiple metabolic markers;

[0025] Extract peak height, peak area, peak position and slit width according to the wavelength in each target wavelength range in the detection result data to obtain characteristic peak parameters;

[0026] Principal component analysis is performed based on the characteristic peak parameters to obtain the metabolic characteristic data.

[0027] In one possible implementation, the multivariate regression algorithm model is constructed as follows:

[0028] Preliminarily constructing a spectrum matrix based on the plurality of identification characteristic absorption peak data, wherein each row in the spectrum matrix represents absorbance data within a preset wavelength range;

[0029] Constructing a concentration matrix in advance based on the sample concentrations of multiple metabolite markers, wherein each row in the concentration matrix corresponds to the concentration of one of the metabolite markers;

[0030] Decomposing the spectrum matrix into latent variables, wherein the latent variables are linear combinations of the original spectra corresponding to the identification characteristic absorption peak data, and are used to explain the concentration change to the greatest extent;

[0031] By minimizing the residual, a linear relationship between the latent variable and the concentration is established, and based on the leave-one-out method, the optimal number of latent variables is determined according to the linear relationship between the latent variable and the concentration, thereby obtaining a multivariate regression algorithm model.

[0032] In a possible implementation, the key metabolic enzyme includes one of the following: melatonin synthesis rate-limiting enzyme, methionine adenosine transferase; the immune effector molecule includes at least one of the following: immunoglobulin A, immune interferon.

[0033] In a possible implementation, the functional amino acid includes one of the following: γ-aminobutyric acid, serine, and tryptophan.

[0034] In one possible implementation, the immune phenotype includes at least one of the following: blood immunoglobulin M content, immunoglobulin G content, interleukin-1β content, intestinal tissue secretory immunoglobulin A content, interleukin-22 content, CD68+ macrophage number, diarrhea rate, and morbidity.

[0035] In conjunction with the second aspect of the present invention, an artificial intelligence-based method for predicting immune function of pigs in an embodiment of the present invention is applied to the prediction and display device described in any one of the first aspects, and the method includes:

[0036] Based on the intestinal flora detection data, the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress is simulated based on a dynamic metabolic network to obtain metabolic immune interaction data, wherein the intestinal flora detection data is obtained by the detection device through detecting the intestinal flora corresponding to multiple omics of the pigs under stress;

[0037] The metabolic immune interaction data is stored in a metabolic immune database, and based on the metabolic signature data and the metabolic immune interaction data stored in the metabolic immune database, a dose-effect relationship between functional amino acids and immune phenotypes is determined based on metabolic flux redirection to obtain metabolic spectrum characteristics of the pig under stress, wherein the metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios, and the metabolic signature data is obtained by detecting test result data of multiple metabolic markers of the pig under stress, and performing feature extraction based on the test result data;

[0038] Based on the metabolic profile characteristics, amino acid ratios or dietary intervention strategies are predicted, generated, and displayed.

[0039] Compared with the existing technology, the pig immune function prediction system and method based on artificial intelligence provided by the present disclosure can achieve at least the following beneficial effects:

[0040] Obtaining intestinal flora detection data corresponding to multiple omics and multiple metabolic marker detection result data of pigs under stress conditions through detection devices helps to understand the physiological status of pigs under stress conditions from multiple dimensions, avoiding the difficulty of capturing the comprehensiveness of metabolic-immune interactions with single metabolomic or immune group data. Feature extraction of the detection result data through the edge optimization decision-making device to obtain metabolic characteristic data helps to focus on key information, remove redundant data, and improve the efficiency and accuracy of subsequent analysis. The prediction and display device simulates the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress conditions based on a dynamic metabolic network, obtains metabolic-immune interaction data, and stores it in the metabolic-immune database. It deeply reveals the intrinsic connection between metabolism and immunity in pigs under stress conditions, and based on the metabolic characteristic data and the metabolic-immune interaction data in the metabolic-immune database, determines the dose-effect relationship between functional amino acids and immune phenotypes through metabolic flow redirection, obtains the metabolic spectrum characteristics of pigs under stress conditions, and predicts and generates amino acid ratios or dietary intervention strategies based on this. It can achieve accurate prediction of the immune function of pigs and formulate scientific and reasonable intervention measures, which will help improve the immunity and health level of pigs, reduce breeding risks and improve breeding efficiency.

[0041] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following detailed description, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0043] Figure 1 This is a block diagram of an artificial intelligence-based pig immune function prediction system provided by an embodiment of the present invention.

[0044] Figure 2 It is a schematic diagram of the execution flow of the method for predicting the immune function of pigs based on artificial intelligence provided by an embodiment of the present invention.

[0045] Figure 3 is a schematic diagram of exemplary hardware and software components of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] The following describes the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure and are not intended to limit the present disclosure.

[0048] The embodiment of the present invention provides a pig immune function prediction system based on artificial intelligence, see Figure 1 As shown, the system includes:

[0049] A detection device for detecting intestinal flora detection data corresponding to multiple omics and detection result data of multiple metabolic markers of the pig under stress;

[0050] In the disclosed embodiment, multiple omics may include metabolomics, immunology, transcriptomics and other omics, and through the fusion analysis framework, the limitations of single omics can be overcome.

[0051] In the disclosed embodiments, the intestinal flora may mainly include enterotoxigenic Escherichia coli. It can be explained that the invasion of enterotoxigenic Escherichia coli can induce the redirection of the metabolic flow of γ-aminobutyric acid (GABA) in the jejunum tissue of pigs, and the GABA signaling pathway mediates the cascade regulation of the post-infection immune response through epigenetic modification. Early weaning stress significantly changes the metabolic profile characteristics of GABA, serine and tryptophan in the serum of piglets. By exogenous supplementation of specific amino acid ratios or dietary intervention strategies, targeted enhancement of the immune barrier function of stress-susceptible pigs can be achieved.

[0052] Metabolic markers can be substances with specific chemical structures and biological activities produced during the metabolic process in piglets under stress. Changes in their levels can reflect the metabolic state and physiological function of the organism. Stress can occur during infection or weaning, revealing the synergistic mechanisms between the amino acid metabolic network, intestinal microbes, and immune cell differentiation under infection or weaning stress.

[0053] In the disclosed embodiments, the detection device utilizes technologies such as high-throughput sequencing, mass spectrometry, and immunological detection to detect intestinal flora and metabolic markers in pigs under stress. High-throughput sequencing can sequence the genes of intestinal flora to analyze their species and abundance; mass spectrometry can detect metabolite levels in samples such as blood and urine to determine the levels of metabolic markers; and immunological detection can detect the concentrations of immune effector molecules produced by immune cells.

[0054] For example, when using high-throughput sequencing to analyze the intestinal microbiome of pigs, DNA is first extracted from fecal samples, and then the 16S rRNA gene is sequenced. By comparing the sequencing results with a known microbiome database, the species and relative abundance of different bacteria in the gut can be determined. For metabolic markers, such as blood glucose levels, a blood sample from the pig can be tested using a glucometer to directly measure blood glucose concentrations.

[0055] An edge optimization decision-making device, configured to extract features based on the detection result data to obtain metabolic feature data;

[0056] In the disclosed embodiments, the edge optimization decision-making device uses machine learning and data mining algorithms to process and analyze the large amount of test result data acquired by the detection device, extracting representative metabolic signature data. These algorithms can automatically identify patterns and regularities in the data, remove noise and redundant information, and transform the raw data into more concise and meaningful signature data.

[0057] For example, suppose a detection device acquires concentration data of multiple metabolites in pig blood. The edge optimization decision-making device can use the principal component analysis (PCA) algorithm to reduce the dimensionality of this data. The PCA algorithm identifies the main components in the data and combines the multiple metabolite concentration data into a few principal components, which are the metabolic signature data. For example, PCA analysis can convert the concentration data of dozens of metabolites into three principal components, which can represent the majority of the information in the original data.

[0058] A prediction and display device is used to simulate the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules of pigs under stress conditions based on a dynamic metabolic network according to the intestinal flora detection data, obtain metabolic immune interaction data, and store the metabolic immune interaction data in a metabolic immune database, and determine the dose-effect relationship between functional amino acids and immune phenotypes based on metabolic flow redirection according to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, so as to obtain the metabolic spectrum characteristics of the pigs under stress conditions. The metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios. According to the metabolic spectrum characteristics, amino acid ratios or dietary intervention strategies are predicted and generated.

[0059] The dynamic metabolic network is used to describe the interconversion relationships between metabolites and the dynamic changes in metabolic reactions in piglets. It takes into account factors such as metabolite concentration, metabolic enzyme activity, and reaction rate, and can simulate the metabolic changes of organisms under different physiological states.

[0060] Metabolic enzymes are enzymes that catalyze metabolic reactions in organisms. They accelerate the conversion and decomposition of metabolites and maintain the normal metabolic function of organisms. Changes in the activity of key metabolic enzymes can affect the operation of the entire metabolic network.

[0061] Immune effector molecules, such as cytokines and antibodies, are molecules produced by immune cells with immunomodulatory effects. They can regulate immune cell activity and promote the onset and progression of immune responses. Metabolic immune interaction data reflect the interaction between metabolic processes and the immune system, including the relationship between metabolic enzyme activity and immune effector molecule expression, and the impact of metabolites on immune cell function.

[0062] Metabolic flux redirection alters the direction and volume of metabolites within an organism, shifting metabolic processes toward specific physiological functions. In predicting pig immune function, metabolic flux redirection can be used to regulate the metabolic pathways of functional amino acids, thereby influencing the immune phenotype. The immune phenotype is the characteristic manifestation of immune cells in terms of morphology, function, and surface markers, reflecting their development, differentiation, and activation status.

[0063] A metabolic profile is the metabolite composition and metabolic characteristics of an organism under a specific physiological state. It comprehensively reflects the metabolic state and physiological function of the organism. Amino acid ratio refers to the ratio of different amino acids in the diet. A reasonable amino acid ratio can meet the nutritional needs of pigs, promote their growth and development, and enhance immune function. Dietary intervention strategies refer to measures that improve the health and immune function of pigs by adjusting the composition and nutritional level of the diet.

[0064] In the disclosed embodiment, the metabolic immune database integrates and stores metabolic-immune interaction data of 12 stress scenarios and 5 types of immune cells, covering the entire life cycle of pigs, and providing a target library for precise intervention.

[0065] In the disclosed embodiments, the prediction and display device constructs a dynamic metabolic network model that encompasses various metabolic reactions and metabolites within the pig body, as well as the interactions between metabolic enzymes and immune effector molecules. By inputting initial conditions such as intestinal flora detection data, mathematical models and computer simulation techniques are used to simulate the changes in the activity of key metabolic enzymes and the expression of immune effector molecules in the pig's metabolic network under stress, thereby generating metabolic-immune interaction data.

[0066] For example, a dynamic metabolic network model encompassing pathways such as sugar metabolism and amino acid metabolism can be constructed. During the simulation, it is assumed that pigs are subjected to transport stress, resulting in changes in the intestinal microbiome and altered concentrations of certain metabolites. These changes are input into the model, which simulates changes in the activity of key metabolic enzymes (such as hexokinase) and the expression levels of immune effector molecules (such as interleukin-6). Analysis of these simulation results reveals the relationship between metabolic enzyme activity and immune effector molecule expression, i.e., metabolic-immune interaction data.

[0067] Furthermore, the obtained metabolic-immune interaction data is organized and stored according to certain formats and standards to establish a metabolic-immune database. The database can be a relational or non-relational database to facilitate data query, management, and update. For example, a MySQL database is used to store metabolic-immune interaction data. The database table structure is designed, including a metabolic enzyme information table, an immune effector molecule information table, a metabolic-immune interaction relationship table, etc. The simulated metabolic enzyme activity, immune effector molecule expression level, and the relationship data between them are inserted into the corresponding tables.

[0068] Furthermore, by leveraging metabolic signature data and data from the metabolic immunity database, combined with metabolic flux analysis and optimization algorithms, the metabolic pathways of functional amino acids were regulated, redirecting metabolic flux toward immune function. By varying the dosage of functional amino acids and observing changes in the immune phenotype, the dose-effect relationship between functional amino acids and immune phenotype was determined. Furthermore, by comprehensively analyzing changes in metabolites, the metabolic profile characteristics of pigs under stress were determined.

[0069] For example, metabolic signature data revealed abnormalities in the metabolism of certain amino acids in pigs under stress. Relevant metabolic-immune interaction data were obtained from the metabolic-immune database to understand the relationship between these amino acids and immune cell function. Then, different doses of functional amino acids (such as arginine) were added to the diet to observe changes in immune phenotypes, such as proliferation and activation, in pig immune cells (such as T cells). Dose-effect curves were fitted to the experimental data to determine the optimal dose of arginine. Simultaneously, metabolite levels in pig blood, tissue, and other samples were measured, and changes in the metabolic profile were analyzed to determine the metabolic profile characteristics of pigs under stress.

[0070] Furthermore, based on the metabolic profiles obtained, machine learning models or expert systems can be used to predict amino acid ratios or dietary intervention strategies that can improve pig immune function, drawing on existing knowledge and experience. These models and systems can automatically generate optimal amino acid ratios or dietary adjustment recommendations based on the relationship between metabolic profiles and immune function.

[0071] For example, a support vector machine (SVM) model is used for prediction. The SVM model is trained using the metabolic profile characteristics of pigs under different stress conditions as input and the corresponding optimal amino acid ratio or effective dietary intervention strategy as output. When a new pig metabolic profile is input, the model can predict the appropriate amino acid ratio or dietary intervention strategy for that pig. For example, it predicts that a 2:1:1 ratio of arginine, glutamine, and threonine in the diet can effectively improve the pig's immune function under transport stress.

[0072] The above technical solution uses a detection device to obtain intestinal flora detection data corresponding to multiple omics and multiple metabolic marker detection result data of pigs under stress conditions, which helps to understand the physiological conditions of pigs under stress conditions from multiple dimensions and avoid the difficulty of capturing the comprehensiveness of metabolic-immune interactions with single metabolomic or immune group data. The edge optimization decision-making device extracts features from the detection result data to obtain metabolic feature data, which helps to focus on key information, remove redundant data, and improve the efficiency and accuracy of subsequent analysis. The prediction and display device simulates the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress conditions based on a dynamic metabolic network, obtains metabolic-immune interaction data, and stores it in the metabolic-immune database. It deeply reveals the intrinsic connection between metabolism and immunity in pigs under stress conditions, and based on the metabolic feature data and the metabolic-immune interaction data in the metabolic-immune database, determines the dose-effect relationship between functional amino acids and immune phenotypes through metabolic flow redirection, obtains the metabolic spectrum characteristics of pigs under stress conditions, and predicts and generates amino acid ratios or dietary intervention strategies based on this. It can achieve accurate prediction of the immune function of pigs and formulate scientific and reasonable intervention measures, which will help improve the immunity and health level of pigs, reduce breeding risks and improve breeding efficiency.

[0073] In one possible implementation, the prediction and display device is specifically used to:

[0074] Based on the composition and abundance of the intestinal flora in the intestinal flora detection data and the metabolites corresponding to the intestinal flora, the nodes in the dynamic metabolic network are deleted to obtain a target metabolic network, wherein the dynamic metabolic network is constructed based on the full amount of metabolites and metabolic enzymes and includes metabolite nodes, metabolic enzyme nodes, and reaction edges, and the reaction edges are used to represent the conversion relationship between metabolites;

[0075] In the disclosed embodiments, the dynamic metabolic network originally covers the full range of metabolites and metabolic enzymes, and is a large and complex network. However, in the actual physiological processes of pigs, not all metabolites and metabolic enzymes are involved in metabolic activities under specific stress states. Intestinal flora detection data provides information about the composition and abundance of intestinal flora and their corresponding metabolites. This information can reflect which metabolites and metabolic enzymes are truly active and relevant under the current stress state.

[0076] Furthermore, by comparing and analyzing the intestinal flora detection data with the dynamic metabolic network, we can identify metabolite nodes and metabolic enzyme nodes that are not related to the metabolic activities of the intestinal flora under the current stress state and delete them from the network, thereby obtaining a target metabolic network that is more in line with the actual stress state and simplifying the complexity of subsequent analysis.

[0077] For example, under normal physiological conditions, the dynamic metabolic network includes numerous metabolites and metabolic enzymes involved in various metabolic pathways, such as fat metabolism, sugar metabolism, and amino acid metabolism. When pigs are under transport stress, tests have found that the abundance of certain short-chain fatty acid-producing bacteria in the intestinal flora has increased significantly, while the activity of some metabolic enzymes involved in fat synthesis is very low, and the corresponding metabolite concentrations remain almost unchanged. Therefore, metabolite nodes related to fat synthesis (such as certain long-chain fatty acids) and metabolic enzyme nodes (such as fatty acid synthase) can be deleted from the dynamic metabolic network to obtain a target metabolic network with short-chain fatty acid metabolism as its main content.

[0078] According to the regulation of the activity of the key metabolic enzyme by the metabolites corresponding to the intestinal flora, the key metabolic enzyme is added from the reaction edge through different metabolic enzyme nodes in the target metabolic network, and the expression of the immune effector molecule by the activity of the key metabolic enzyme of the pig under stress is simulated to obtain the correlation coefficient between the activity of the key metabolic enzyme and the expression of the immune effector molecule under different stress conditions;

[0079] In the disclosed embodiments, the metabolites produced by the intestinal flora can regulate the activity of key metabolic enzymes. In the target metabolic network, different metabolic enzyme nodes represent different metabolic enzymes. By adding key metabolic enzymes from the reaction side, the changes in the activity of key metabolic enzymes due to the regulation of intestinal flora metabolites under stress can be simulated. The expression of immune effector molecules is closely related to the activity of key metabolic enzymes. Changes in the activity of key metabolic enzymes will affect the generation of metabolites and the direction of metabolic pathways, thereby affecting the activation of immune cells and the production of immune effector molecules.

[0080] By simulating the changes in the activity of key metabolic enzymes under different stress conditions and observing the changes in the expression of corresponding immune effector molecules, statistical methods (such as Pearson correlation coefficient calculation) can be used to obtain the correlation coefficient between the activity of key metabolic enzymes and the expression of immune effector molecules under different stress conditions, thereby quantifying the relationship between them.

[0081] In the target metabolic network, butyrate produced by intestinal microbiota is known to regulate the activity of histone deacetylase (HDAC). To simulate transport stress, the amount of HDAC added to the reaction was gradually increased (simulating enhanced HDAC activity), while the expression of the immune effector molecule interleukin-10 (IL-10) was simultaneously monitored. It was found that with increasing HDAC addition, IL-10 expression also showed an upward trend. Pearson correlation coefficient calculation showed a correlation coefficient of 0.8 between HDAC activity and IL-10 expression under transport stress conditions, indicating a strong positive correlation between them.

[0082] A dynamic model based on flux balance analysis is used in combination with a two-level optimization algorithm to simulate changes in metabolic flux under stress conditions, and a mixed integer optimization method is used to extract the minimum equilibrium pathway from substrates to the metabolites to optimize the target metabolic network.

[0083] In the disclosed embodiments, a dynamic model based on flux balance analysis can simulate the distribution and changes of metabolic fluxes in a metabolic network. It takes into account factors such as metabolite concentrations, metabolic enzyme activity, and reaction rates. It uses optimization algorithms (such as a two-level optimization algorithm) to find the optimal distribution of metabolic fluxes under stress, enabling the metabolic network to more accurately reflect actual metabolic processes. Mixed integer optimization methods are used to extract the minimum equilibrium pathway from substrates to metabolites, specifically to find the simplest and most efficient metabolic pathway while meeting metabolic requirements.

[0084] Through these optimization methods, some unnecessary reaction edges and nodes in the target metabolic network can be removed, the direction and rate of metabolic reactions can be corrected, and the target metabolic network can be made more reasonable and efficient, and more in line with the actual metabolic conditions of pigs under stress.

[0085] Taking the sugar metabolism pathway within the target metabolic network as an example, a dynamic model based on flux balance analysis, combined with a two-level optimization algorithm, was used to simulate the metabolic flow of glucose within the metabolic network under transport stress. It was discovered that some irrational metabolic branches in the original network resulted in inefficient glucose metabolism. Using a mixed integer optimization method, the minimal equilibrium pathway from glucose to substrates (such as pyruvate) was extracted. Unnecessary branch reactions were removed, and the activity parameters of key metabolic enzymes were adjusted, resulting in more efficient conversion of glucose to pyruvate and its further participation in subsequent metabolic processes, ultimately optimizing the target metabolic network.

[0086] Based on the optimized target metabolic network, re-performing the step of adding key metabolic enzymes from the reaction edge through different metabolic enzyme nodes in the target metabolic network, simulating the effect of the activity of the key metabolic enzymes on the expression of the immune effector molecules under stress conditions, and obtaining correlation coefficients between the activity of the key metabolic enzymes and the expression of the immune effector molecules under different stress conditions, until the difference between the correlation coefficients obtained from two consecutive calculations is less than a preset coefficient threshold;

[0087] In the disclosed embodiment, the target metabolic network is optimized, and the optimized network can more accurately reflect the metabolic status of pigs under stress. Therefore, it is necessary to re-execute the simulation process, add key metabolic enzymes from the reaction edge through different metabolic enzyme nodes in the target metabolic network, simulate the effect of the activity of key metabolic enzymes on the expression of immune effector molecules, and recalculate the correlation coefficient. Compare the correlation coefficient obtained in this calculation with the correlation coefficient obtained in the previous calculation. If the difference between them is greater than the preset coefficient threshold, it means that the optimized network has a significant impact on the simulation results, and it is necessary to continue to optimize the network and re-simulate the calculation; if the difference is less than the preset coefficient threshold, it means that the simulation results have stabilized, the network optimization has achieved a good effect, and the iteration can be stopped.

[0088] Assume that the preset coefficient threshold is 0.05. After the first optimization of the target metabolic network, the relationship between HDAC activity and IL-10 expression under transport stress is re-simulated, resulting in a correlation coefficient of 0.82. Compared to the previously calculated correlation coefficient of 0.8, the difference is 0.02, which is less than the preset coefficient threshold of 0.05. At this point, iterations can be stopped, assuming that the current target metabolic network and simulation results are sufficiently representative of the actual situation.

[0089] The metabolic immune interaction data is determined based on the addition amount of the key metabolic enzyme corresponding to the correlation coefficient obtained in the last calculation and the expression amount of the corresponding immune effector molecule.

[0090] In the disclosed examples, the correlation coefficients obtained from the final calculation correspond to the quantitative relationship between the addition level (i.e., activity) of key metabolic enzymes and the expression level of immune effector molecules under the optimal target metabolic network. These data comprehensively and accurately reflect the interaction between metabolic processes and the immune system in pigs under stress, and thus can be used as metabolic-immune interaction data. These data are of great reference value for deepening the understanding of the physiological mechanisms of pigs under stress and formulating targeted breeding management measures and nutritional regulation strategies.

[0091] Taking the above example as an example, according to the last calculation, under the transport stress state, when the amount (activity) of HDAC added reaches a certain level, the expression level of IL-10 also reaches the corresponding peak, and the correlation coefficient between them is 0.82. The data such as the amount of HDAC added, the expression level of IL-10 and the correlation coefficient between them are sorted out to form the metabolic immune interaction data between HDAC and IL-10 under the transport stress state.

[0092] In one possible implementation, the prediction and display device is specifically used to:

[0093] According to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, based on metabolic flux redirection, computer simulation technology is used to predict the dynamic changes of metabolic flux to obtain the concentration change data of the metabolite;

[0094] In the disclosed embodiments, the metabolic flux in the piglets is not fixed, but is different under different stress states, and can be adjusted according to changes in physiological state and external environment. Metabolic flux redirection refers to changing the activity of key metabolic enzymes in metabolic pathways, the concentration of metabolites and other factors, so that the metabolic flux develops in a direction that is beneficial to specific physiological functions. When pigs are under stress, the relationship between intestinal flora and metabolic immune interactions changes, which affects the balance of the metabolic network and causes the metabolic flux to redirect.

[0095] Computer simulation techniques can be used to construct metabolic network models that encompass various metabolites, metabolic enzymes, and the reaction relationships between them. By inputting metabolic signature data (such as initial metabolite concentrations and metabolic enzyme activities) and metabolic-immune interaction data stored in the metabolic-immune database (such as the relationship between metabolic enzyme activity and the impact of immune effector molecule expression on metabolism), the dynamic changes in metabolic flux within the metabolic network can be simulated. Based on predefined algorithms and rules, computer simulations can calculate the production and consumption rates of metabolites at each time point, thereby generating data on changes in metabolite concentrations.

[0096] The hypothesis is that under weaning stress, elevated concentrations of certain metabolites (such as short-chain fatty acids) produced by the gut microbiota of pigs are detected. These metabolites may affect the activity of key metabolic enzymes in the sugar metabolism pathway. Based on information from the Metabolic Immunity Database, these short-chain fatty acids are known to activate hexokinase in the glycolysis pathway.

[0097] A metabolic network model encompassing the sugar metabolism pathway was constructed using computer simulation technology. Input data included the initial glucose concentration, the initial hexokinase activity, and the concentration changes of short-chain fatty acids. During the simulation, as the short-chain fatty acid concentration increased, hexokinase activity increased, metabolic flux in the glycolysis pathway increased, glucose consumption accelerated, and metabolites such as pyruvate were generated. By calculating the changes in glucose and pyruvate concentrations at each time point, their concentration change data were obtained. For example, during the simulation period of 0 to 12 hours, the glucose concentration gradually decreased from an initial 5 mmol / L to 3 mmol / L, while the pyruvate concentration gradually increased from 0 mmol / L to 1.5 mmol / L.

[0098] Predicting the metabolite accumulation data and functional amino acid consumption data of the pig under the stress state based on the metabolic flux and the metabolite concentration change data;

[0099] In the disclosed embodiments, metabolic flux describes the direction and rate of metabolite flow in a metabolic network. By analyzing the direction and magnitude of metabolic flux, it is possible to determine which metabolites accumulate in a particular pathway and which metabolites are rapidly consumed. For example, if a metabolic pathway has a large metabolic flux and a slow downstream reaction rate, then upstream metabolites may accumulate in that pathway.

[0100] Continuing with the example of the sugar metabolism pathway, the metabolic flux and metabolite concentration change data revealed that the glycolysis pathway has a high metabolic flux, while the rate at which pyruvate enters the tricarboxylic acid cycle is relatively slow. Therefore, it is predicted that pyruvate will accumulate in the glycolysis pathway. By calculating the difference between the amount of pyruvate produced and the amount consumed by the tricarboxylic acid cycle over a simulated 12-hour period, the pyruvate accumulation data was obtained, for example, an accumulation of 0.8 mmol / L.

[0101] Functional amino acids participate in various physiological processes in pig metabolism, including protein synthesis, energy metabolism, and immune regulation. Under stress, the pig's demand for and consumption of functional amino acids can change. By analyzing the reactions involving functional amino acids in metabolic flux and combining them with metabolite concentration data, functional amino acid consumption can be calculated. Furthermore, based on the direction of metabolic flux and metabolite accumulation, the amount of metabolite accumulation can be predicted.

[0102] Functional amino acids, such as glutamine, are involved in immune cell energy supply and protein synthesis under stress. Analysis of glutamine-related reactions in metabolic flux revealed its involvement in glutathione synthesis and immune cell energy supply. Combined with metabolite concentration data, glutamine consumption was calculated to be 2 mmol / L over a 12-hour period.

[0103] Based on the accumulation data of metabolites and the consumption data of functional amino acids under the stress state, the dose-effect relationship between the functional amino acids and the immune phenotype is determined to obtain the metabolic profile characteristics of the pigs under the stress state.

[0104] In the disclosed embodiments, there is a certain dose-effect relationship between the consumption of functional amino acids and the immune phenotype. The immune phenotype reflects the development, differentiation, and activation state of immune cells, such as the proliferation capacity of immune cells and the secretion level of cytokines. By analyzing the changes in the immune phenotype under different dosages of functional amino acid consumption, the dose-effect relationship between them can be determined. For example, within a certain range, as the consumption of functional amino acids increases, the proliferation capacity of immune cells may be enhanced, but when the consumption exceeds a certain threshold, the function of immune cells may be suppressed.

[0105] Different doses of glutamine were added to the diet to observe changes in the immune phenotype of pigs under transport stress, such as the proliferation of T lymphocytes in peripheral blood. They found that T lymphocyte proliferation was strongest when glutamine consumption was between 1 and 1.5 mmol / L; however, T lymphocyte proliferation decreased when consumption was below 1 mmol / L or above 1.5 mmol / L. This established a dose-response relationship between glutamine consumption and T lymphocyte proliferation.

[0106] A metabolic profile refers to the metabolite composition and metabolic characteristics of an organism under a specific physiological state. It comprehensively reflects the organism's metabolic state and physiological function. Based on metabolite accumulation data and the dose-response relationship between functional amino acids and immune phenotypes, a comprehensive analysis of pig metabolism under stress can be used to generate a metabolic profile. Metabolic profiles can include information such as metabolite types and concentrations, metabolic pathway activity, and changes in immune function.

[0107] Continuing with the above example, we obtained data on pyruvate accumulation (0.8 mmol / L), as well as the accumulation of other metabolites and the consumption of functional amino acids, to comprehensively analyze the metabolic status of pigs under transport stress. We found that in addition to pyruvate accumulation, the metabolism of certain amino acids also changed, such as increased arginine consumption.

[0108] Combining this information revealed metabolic profiles of pigs under transport stress, including active glucose metabolism pathways, pyruvate accumulation, changes in glutamine and arginine metabolism, and a specific dose-response relationship between T lymphocyte proliferation and glutamine consumption. These metabolic profiles provide a valuable basis for understanding the physiological mechanisms of pigs under stress and developing nutritional regulation strategies.

[0109] In one possible implementation, the detection device includes a near-infrared rapid detection module configured to:

[0110] determining infrared band data for spectrum acquisition based on characteristic absorption peaks of chemical bonds in the plurality of metabolic markers;

[0111] Among them, different chemical bonds have specific vibration and rotation frequencies, and these frequencies correspond to specific infrared absorption wavelengths. In infrared spectroscopy, chemical bonds absorb infrared light of specific wavelengths, thereby producing characteristic absorption peaks. Metabolic markers are molecules composed of various chemical bonds. By studying the chemical structure of these metabolite markers, we can determine the type of chemical bonds they contain, and then find the wavelength range where the characteristic absorption peaks corresponding to these chemical bonds are located. Determining the appropriate infrared band data is the basis for accurate spectral acquisition, because only by selecting the band containing the characteristic absorption peaks of metabolite markers can these metabolite markers be effectively detected in subsequent spectral acquisition and analysis.

[0112] Assume that one of the metabolic markers to be detected is glucose. Glucose molecules contain chemical bonds such as CH, OH, and CO. Among them, the stretching vibration absorption peak of the CH bond is usually between 2800 and 3000 cm -1 Band, the stretching vibration absorption peak of OH bond is at 3200-3600cm -1 The stretching vibration absorption peak of CO bond is between 1000-1300 cm -1 Band. Through the study of the characteristic absorption peaks of glucose chemical bonds, it was determined that the infrared band data used for spectrum acquisition is 2800-3600cm -1 and 1000-1300cm -1 , which ensures that the characteristic absorption information of glucose molecules can be captured during the subsequent spectrum acquisition process.

[0113] According to the infrared band data, spectrum acquisition is performed on the collected blood sample of the live pig to obtain a scanning spectrum;

[0114] In the disclosed embodiments, the parameters of a near-infrared spectrometer are set based on the determined infrared band data, so that it emits infrared light in the corresponding band. When infrared light is irradiated onto a pig blood sample, various components in the blood (including metabolic markers) absorb the infrared light to varying degrees, resulting in variations in the intensity of the infrared light transmitted through the sample. The spectrometer detects the intensity of the infrared light transmitted through the sample, converts it into an electrical signal, and records it to produce a scanned spectrum. This scanned spectrum reflects the absorption behavior of the blood sample at different wavelengths and contains spectral information about the various components in the blood.

[0115] Using a near-infrared spectrometer, the wavelength of 2800-3600 cm-1 and 1000-1300cm -1 The infrared band data sets the instrument parameters. The collected pig blood sample is placed in the sample pool of the spectrometer, and the instrument emits infrared light of the corresponding band to illuminate the sample. Metabolic markers such as glucose, amino acids, and other components in the blood absorb the infrared light. The infrared light that passes through the sample is detected and recorded by the spectrometer, resulting in a scanning spectrum curve. This curve is between 2800-3000cm -1 、3200-3600cm -1 and 1000-1300cm -1 There will be absorption peaks in the same band, which correspond to the absorption of infrared light by different components in the blood.

[0116] performing characteristic absorption peak identification of the plurality of metabolic markers on the scanned spectrum to obtain a plurality of identification characteristic absorption peak data;

[0117] In the disclosed embodiments, each metabolite marker has its own unique chemical structure and chemical bond combination, and therefore exhibits a specific characteristic absorption peak in the infrared spectrum. By studying and analyzing the spectral characteristics of known metabolite markers, a database of characteristic absorption peaks of metabolite markers is established. The obtained scanning spectrum is compared and matched with the characteristic absorption peaks in the database to identify the characteristic absorption peaks belonging to multiple metabolite markers in the scanning spectrum. The process of identifying characteristic absorption peaks needs to consider factors such as the position, intensity, and shape of the absorption peak to ensure accurate identification of the spectral characteristics of the metabolite marker. The identified characteristic absorption peak data contains specific information about the metabolite marker in the spectrum, such as the wavelength and intensity of the absorption peak.

[0118] The multiple identification characteristic absorption peak data are substituted into a multivariate regression algorithm model, which is used to process the relationship between highly collinear spectral data and concentration to obtain detection result data of multiple metabolic markers.

[0119] In the disclosed embodiment, the near-infrared rapid detection module can be controlled based on the microfluidic chip to achieve real-time detection. The detection device integrates the microfluidic chip and the near-infrared rapid detection module to complete the detection of various metabolic markers such as glutamine and γ-aminobutyric acid in the blood within 15 minutes, and warn of the risk of immunosuppression caused by weaning stress.

[0120] In the disclosed embodiments, in near-infrared spectral analysis, due to the high collinearity of spectral data (i.e., strong correlations between spectral data at different wavelengths), traditional univariate analysis methods have difficulty accurately establishing the relationship between spectral data and metabolic marker concentrations. Multivariate regression algorithm models (such as partial least squares regression and principal component regression) are capable of handling the relationship between highly collinear spectral data and concentrations. This model establishes a mathematical model between spectral characteristic absorption peak data and metabolic marker concentrations by learning and training on the spectral data of a large number of metabolic marker samples with known concentrations. The resulting multiple identification characteristic absorption peak data are substituted into the trained multivariate regression algorithm model. The model calculates the concentrations of multiple metabolic markers based on the pre-established mathematical relationship, thereby obtaining the test result data.

[0121] In the disclosed embodiment, a partial least squares regression algorithm is used to establish a model. First, a large number of pig blood samples with known glucose and lysine concentrations are collected, and near-infrared spectroscopy is performed on these samples to obtain spectral data, and the characteristic absorption peak data of glucose and lysine are identified. Then, these characteristic absorption peak data and the corresponding metabolic marker concentrations are used as training data to train the partial least squares regression model. After the training is completed, the characteristic absorption peak data of glucose and lysine of the unknown samples are substituted into the trained model. The model performs calculations based on the pre-established mathematical relationship to obtain the detection result data of multiple metabolic markers.

[0122] In a possible implementation, the edge optimization decision-making device is specifically configured to:

[0123] determining target wavelength ranges for the multiple metabolic markers based on the vibration characteristics of the chemical bonds in the multiple metabolic markers;

[0124] In the disclosed embodiments, the vibration characteristics of chemical bonds determine their absorption wavelengths in the infrared spectrum. When chemical bonds in molecules (such as C-H, O-H, C=O, etc.) absorb infrared light of a specific wavelength, a transition of vibrational energy levels will occur. Different chemical bonds have different vibration frequencies, and the corresponding absorption wavelengths are also different. Metabolic markers are molecules composed of specific chemical bonds. By studying the chemical structure of these metabolic markers, the types of chemical bonds they contain can be determined, and then, based on the relationship between the vibration frequency and wavelength of the chemical bonds, the absorption wavelength range corresponding to these chemical bonds can be found, that is, the target wavelength range. Determining the target wavelength range is the basis for the subsequent accurate extraction of characteristic peak parameters, because only by selecting a wavelength range that contains the characteristic absorption of the metabolic marker can the spectral information related to the metabolic marker be effectively extracted.

[0125] Extract peak height, peak area, peak position and slit width according to the wavelength in each target wavelength range in the detection result data to obtain characteristic peak parameters;

[0126] In the disclosed embodiments, in spectral analysis, characteristic peak parameters are important indicators for describing the characteristics of spectral peaks. Peak height reflects the intensity of the spectral peak, that is, the degree of absorption of light by the substance at this wavelength; peak area is related to the concentration of the substance. Under certain conditions, the larger the peak area, the higher the concentration of the substance may be; the peak position corresponds to the vibration frequency of the chemical bond, and different chemical bonds or molecular structures will produce absorption peaks at specific wavelength positions; the slit width is related to the shape and resolution of the spectral peak. These characteristic peak parameters can be extracted by processing and analyzing the spectral data within the target wavelength range. The process of extracting characteristic peak parameters requires the use of specific algorithms and software tools, such as smoothing the spectral data to reduce noise interference, and then finding the position of the spectral peak through the peak-finding algorithm, and calculating parameters such as peak height, peak area and slit width.

[0127] Principal component analysis is performed based on the characteristic peak parameters to obtain the metabolic characteristic data.

[0128] In the disclosed embodiment, the edge optimization decision-making device can be a lightweight edge computing terminal developed to support local metabolic feature extraction and dynamic update of cloud models, with a response time of <5 seconds, thereby achieving precise delivery of probiotic preparations and amino acid supplements.

[0129] In the embodiment of the present disclosure, a large number of characteristic peak parameters are obtained in the spectral analysis. There may be a certain correlation between these parameters, resulting in a high data dimension and information redundancy. Principal component analysis converts the original characteristic peak parameters into a set of new, unrelated comprehensive variables, namely principal components, through linear transformation. These principal components are arranged in order of decreasing variance. The first principal component contains the largest variance information in the original data, the second principal component contains the second largest variance information, and so on. By selecting the first few principal components, the dimension of the data can be reduced while retaining most of the data information to obtain metabolic characteristic data. Metabolic characteristic data can more concisely reflect the characteristic information of metabolic markers in the sample.

[0130] In one possible implementation, the multivariate regression algorithm model is constructed as follows:

[0131] Preliminarily constructing a spectrum matrix based on the plurality of identification characteristic absorption peak data, wherein each row in the spectrum matrix represents absorbance data within a preset wavelength range;

[0132] In the embodiment of the present disclosure, in the spectral analysis, each sample will have different absorbance values ​​within a specific wavelength range, and these absorbance values ​​reflect the absorption of light by the substance in the sample. For multiple samples, each sample will have a series of absorbance data within a preset wavelength range. These absorbance data are arranged in a certain order, with each row representing the absorbance data of a sample within a preset wavelength range, and a matrix, namely a spectral matrix, can be constructed. The spectral matrix can systematically organize and manage the spectral information of different samples, providing a basis for subsequent data analysis and model building. By constructing a spectral matrix, the spectral data of different samples can be easily compared and analyzed.

[0133] Constructing a concentration matrix in advance based on the sample concentrations of multiple metabolite markers, wherein each row in the concentration matrix corresponds to the concentration of one of the metabolite markers;

[0134] In the disclosed embodiments, in the detection of metabolic markers, the concentration of different metabolic markers in each sample is an important analytical indicator. For multiple samples, each sample may contain multiple metabolic markers, and each metabolic marker has a corresponding concentration value. These concentration values ​​are arranged in a certain order, with each row corresponding to the concentration of a metabolic marker, and a matrix, i.e., a concentration matrix, can be constructed. The concentration matrix can clearly represent the concentration information of various metabolic markers in different samples, corresponding to the spectral matrix, and facilitates the subsequent establishment of a relationship model between spectral data and concentration. By constructing a concentration matrix, the concentration differences of metabolic markers in different samples can be intuitively compared.

[0135] Decomposing the spectrum matrix into latent variables, wherein the latent variables are linear combinations of the original spectra corresponding to the identification characteristic absorption peak data, and are used to explain the concentration change to the greatest extent;

[0136] In the disclosed embodiment, there is a strong correlation between the spectral data at different wavelengths, which makes it difficult to directly establish the relationship between the spectral data and the concentration. Decomposing the spectral matrix into latent variables is an effective dimensionality reduction and feature extraction method. The latent variable is a linear combination of the original spectra, which can capture the main information in the spectral data while reducing the redundancy between the data. By finding suitable latent variables, the changes in concentration can be explained to the greatest extent, that is, there is a strong correlation between the latent variables and the concentration of metabolic markers. In this way, fewer latent variables can be used to represent the original spectral data, simplifying the structure of the model and improving the stability and predictive ability of the model.

[0137] By minimizing the residual, a linear relationship between the latent variable and the concentration is established, and based on the leave-one-out method, the optimal number of latent variables is determined according to the linear relationship between the latent variable and the concentration, thereby obtaining a multivariate regression algorithm model.

[0138] In the disclosed embodiments, establishing a linear relationship between latent variables and concentration is the core of the multivariate regression algorithm model. By minimizing the residual error (i.e., the difference between the predicted concentration and the actual concentration), the coefficient of the linear relationship can be determined, ensuring that the model's predictions are as close to the true values ​​as possible. When determining the optimal number of latent variables, a balance must be struck between model complexity and predictive power. If the number of latent variables is too small, the model may not fully capture the relationship between spectral data and concentration, resulting in low prediction accuracy. If the number of latent variables is too large, the model may overfit, meaning it performs well on the training data but has poor predictive power on new data. The leave-one-out cross-validation method is a commonly used cross-validation technique. It uses one sample from a sample set as the test set and the remaining samples as the training set, repeating this process until each sample has been used as the test set at least once. By comparing the model's predictive performance (e.g., root mean square error (RMSE) and coefficient of determination (R²)) with different numbers of latent variables, the number of latent variables with the best predictive performance is selected as the optimal number, thereby obtaining the final multivariate regression algorithm model.

[0139] In a possible implementation, the key metabolic enzyme includes one of the following: melatonin synthesis rate-limiting enzyme, methionine adenosine transferase; the immune effector molecule includes at least one of the following: immunoglobulin A, immune interferon.

[0140] In a possible implementation, the functional amino acid includes one of the following: γ-aminobutyric acid, serine, and tryptophan.

[0141] In one possible implementation, the immune phenotype includes at least one of the following: blood immunoglobulin M content, immunoglobulin G content, interleukin-1β content, intestinal tissue secretory immunoglobulin A content, interleukin-22 content, CD68+ macrophage number, diarrhea rate, and morbidity.

[0142] In the disclosed embodiments, machine learning was used to establish a dose-effect relationship between a variety of functional amino acids and their derivatives (γ-aminobutyric acid, serine, and tryptophan) and immune phenotypes (such as blood IgM, IgG, and IL-1β levels, intestinal tissue SIgA and IL-22 levels, CD68+ macrophage numbers, diarrhea rate, and morbidity) to optimize the amino acid supplementation threshold under low-protein diet conditions.

[0143] Based on experimental verification of the system provided by the embodiments of the present disclosure, the diarrhea rate of weaned piglets was reduced by 61.8%, the survival rate was increased by 4.0%, and the daily weight gain throughout the process was increased by 7.1%; the use of antibiotics was reduced by 38.7%, promoting the transformation of pig farming to a green and low-carbon one.

[0144] The embodiment of the present invention provides a method for predicting the immune function of pigs based on artificial intelligence, which is applied to the prediction and display device described in any of the above embodiments, see Figure 2 As shown, the method includes:

[0145] In step S21, based on the intestinal flora detection data, the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in the pig under stress is simulated based on a dynamic metabolic network to obtain metabolic immune interaction data, wherein the intestinal flora detection data is obtained by the detection device by detecting the intestinal flora corresponding to multiple omics of the pig under stress;

[0146] In step S22, the metabolic immune interaction data is stored in a metabolic immune database, and based on the metabolic signature data and the metabolic immune interaction data stored in the metabolic immune database, a dose-effect relationship between functional amino acids and immune phenotypes is determined based on metabolic flux redirection to obtain metabolic spectrum characteristics of the pig under stress. The metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios. The metabolic signature data is obtained by detecting test result data of multiple metabolic markers of the pig under stress and performing feature extraction based on the test result data.

[0147] In step S23, based on the metabolic profile characteristics, an amino acid ratio or a dietary intervention strategy is predicted, generated, and displayed.

[0148] In one possible implementation, in step S21, based on the intestinal flora detection data, the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress is simulated based on a dynamic metabolic network to obtain metabolic immune interaction data, including:

[0149] Based on the composition and abundance of the intestinal flora in the intestinal flora detection data and the metabolites corresponding to the intestinal flora, the nodes in the dynamic metabolic network are deleted to obtain a target metabolic network, wherein the dynamic metabolic network is constructed based on the full amount of metabolites and metabolic enzymes and includes metabolite nodes, metabolic enzyme nodes, and reaction edges, and the reaction edges are used to represent the conversion relationship between metabolites;

[0150] According to the regulation of the activity of the key metabolic enzyme by the metabolites corresponding to the intestinal flora, the key metabolic enzyme is added from the reaction edge through different metabolic enzyme nodes in the target metabolic network, and the expression of the immune effector molecule by the activity of the key metabolic enzyme of the pig under stress is simulated to obtain the correlation coefficient between the activity of the key metabolic enzyme and the expression of the immune effector molecule under different stress conditions;

[0151] A dynamic model based on flux balance analysis is used in combination with a two-level optimization algorithm to simulate changes in metabolic flux under stress conditions, and a mixed integer optimization method is used to extract the minimum equilibrium pathway from substrates to the metabolites to optimize the target metabolic network.

[0152] Based on the optimized target metabolic network, re-performing the step of adding key metabolic enzymes from the reaction edge through different metabolic enzyme nodes in the target metabolic network, simulating the effect of the activity of the key metabolic enzymes on the expression of the immune effector molecules under stress conditions, and obtaining correlation coefficients between the activity of the key metabolic enzymes and the expression of the immune effector molecules under different stress conditions, until the difference between the correlation coefficients obtained from two consecutive calculations is less than a preset coefficient threshold;

[0153] The metabolic immune interaction data is determined based on the addition amount of the key metabolic enzyme corresponding to the correlation coefficient obtained in the last calculation and the expression amount of the corresponding immune effector molecule.

[0154] In one possible implementation, in step S22, determining the dose-effect relationship between functional amino acids and immune phenotypes based on metabolic flux redirection according to the metabolic signature data and the metabolic-immune interaction data stored in the metabolic-immune database, and obtaining the metabolic profile characteristics of the pig under stress state include:

[0155] According to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, based on metabolic flux redirection, computer simulation technology is used to predict the dynamic changes of metabolic flux to obtain the concentration change data of the metabolite;

[0156] Predicting the metabolite accumulation data and functional amino acid consumption data of the pig under the stress state based on the metabolic flux and the metabolite concentration change data;

[0157] Based on the accumulation data of metabolites and the consumption data of functional amino acids under the stress state, the dose-effect relationship between the functional amino acids and the immune phenotype is determined to obtain the metabolic profile characteristics of the pigs under the stress state.

[0158] In a possible implementation, the key metabolic enzyme includes one of the following: melatonin synthesis rate-limiting enzyme, methionine adenosine transferase; the immune effector molecule includes at least one of the following: immunoglobulin A, immune interferon.

[0159] In a possible implementation, the functional amino acid includes one of the following: γ-aminobutyric acid, serine, and tryptophan.

[0160] In one possible implementation, the immune phenotype includes at least one of the following: blood immunoglobulin M content, immunoglobulin G content, interleukin-1β content, intestinal tissue secretory immunoglobulin A content, interleukin-22 content, CD68+ macrophage number, diarrhea rate, and morbidity.

[0161] The implementation methods of the steps in the disclosed method can be derived from the implementation examples of the system and will not be described in detail here.

[0162] Figure 3 The electronic device 100 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the electronic device 100 can be configured as a prediction and display device as described in any of the aforementioned embodiments. The electronic device 100 may also include a communication component 1004, which can be used for data interaction between the device 100 and other devices, such as sending data and / or receiving data. It should be noted that in actual scheduling, the communication component 1004 is not limited to one, and the structure of the electronic device 100 does not constitute a limitation on the embodiments of the present application.

[0163] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0164] Bus 1002 may include a path for transmitting information between the above components. Bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0165] The memory 1003 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0166] The memory 1003 is used to store program codes for executing the embodiments of the present disclosure, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the embodiment of the above-mentioned method for predicting the immune function of pigs based on artificial intelligence.

[0167] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, various changes, modifications, replacements and variations can be made to these embodiments, and these changes, modifications, replacements and variations all fall within the scope of protection of the present disclosure.

[0168] It should also be noted that the various specific technical features described in the above specific embodiments may be combined in any suitable manner, unless there is any contradiction, and these combinations shall also be considered as the contents disclosed in this disclosure. To avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents of the specification and must be determined based on the scope of the claims.

Claims

1. A pig immune function prediction system based on artificial intelligence, characterized in that: The system comprises: A detection device for detecting intestinal flora detection data corresponding to multiple omics and detection result data of multiple metabolic markers of the pig under stress; An edge optimization decision-making device, configured to extract features based on the detection result data to obtain metabolic feature data; A prediction and display device is used to simulate the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress conditions based on a dynamic metabolic network according to the intestinal flora detection data, obtain metabolic immune interaction data, and store the metabolic immune interaction data in a metabolic immune database, and determine the dose-effect relationship between functional amino acids and immune phenotypes based on metabolic flow redirection according to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, so as to obtain the metabolic spectrum characteristics of the pigs under stress conditions, wherein the metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios, and predicts, generates and displays amino acid ratios or dietary intervention strategies based on the metabolic spectrum characteristics.

2. The pig immune function prediction system based on artificial intelligence according to claim 1, characterized in that: The prediction and display device is specifically used for: Based on the composition and abundance of the intestinal flora in the intestinal flora detection data and the metabolites corresponding to the intestinal flora, the nodes in the dynamic metabolic network are deleted to obtain a target metabolic network, wherein the dynamic metabolic network is constructed based on the full amount of metabolites and metabolic enzymes and includes metabolite nodes, metabolic enzyme nodes, and reaction edges, and the reaction edges are used to represent the conversion relationship between metabolites; According to the regulation of the activity of the key metabolic enzyme by the metabolites corresponding to the intestinal flora, the key metabolic enzyme is added from the reaction edge through different metabolic enzyme nodes in the target metabolic network, and the expression of the immune effector molecule by the activity of the key metabolic enzyme of the pig under stress is simulated to obtain the correlation coefficient between the activity of the key metabolic enzyme and the expression of the immune effector molecule under different stress conditions; A dynamic model based on flux balance analysis is used in combination with a two-level optimization algorithm to simulate changes in metabolic flux under stress conditions, and a mixed integer optimization method is used to extract the minimum equilibrium pathway from substrates to the metabolites to optimize the target metabolic network. Based on the optimized target metabolic network, re-performing the step of adding key metabolic enzymes from the reaction edge through different metabolic enzyme nodes in the target metabolic network, simulating the effect of the activity of the key metabolic enzymes on the expression of the immune effector molecules under stress conditions, and obtaining correlation coefficients between the activity of the key metabolic enzymes and the expression of the immune effector molecules under different stress conditions, until the difference between the correlation coefficients obtained from two consecutive calculations is less than a preset coefficient threshold; The metabolic immune interaction data is determined based on the addition amount of the key metabolic enzyme corresponding to the correlation coefficient obtained in the last calculation and the expression amount of the corresponding immune effector molecule.

3. The pig immune function prediction system based on artificial intelligence according to claim 1, characterized in that: The prediction and display device is specifically used for: According to the metabolic characteristic data and the metabolic immune interaction data stored in the metabolic immune database, based on metabolic flux redirection, computer simulation technology is used to predict the dynamic changes of metabolic flux to obtain the concentration change data of the metabolite; Predicting the metabolite accumulation data and functional amino acid consumption data of the pig under the stress state based on the metabolic flux and the metabolite concentration change data; Based on the accumulation data of metabolites and the consumption data of functional amino acids under the stress state, the dose-effect relationship between the functional amino acids and the immune phenotype is determined to obtain the metabolic profile characteristics of the pigs under the stress state.

4. The pig immune function prediction system based on artificial intelligence according to claim 1, characterized in that: The detection device includes a near-infrared rapid detection module for: determining infrared band data for spectrum acquisition based on characteristic absorption peaks of chemical bonds in the plurality of metabolic markers; According to the infrared band data, spectrum acquisition is performed on the collected blood sample of the live pig to obtain a scanning spectrum; performing characteristic absorption peak identification of the plurality of metabolic markers on the scanned spectrum to obtain a plurality of identification characteristic absorption peak data; The multiple identification characteristic absorption peak data are substituted into a multivariate regression algorithm model, which is used to process the relationship between highly collinear spectral data and concentration to obtain detection result data of multiple metabolic markers.

5. The pig immune function prediction system based on artificial intelligence according to claim 4 is characterized in that: The edge optimization decision-making device is specifically used to: determining target wavelength ranges for the multiple metabolic markers based on the vibration characteristics of the chemical bonds in the multiple metabolic markers; Extract peak height, peak area, peak position and slit width according to the wavelength in each target wavelength range in the detection result data to obtain characteristic peak parameters; Principal component analysis is performed based on the characteristic peak parameters to obtain the metabolic characteristic data.

6. The pig immune function prediction system based on artificial intelligence according to claim 4 is characterized in that: The multivariate regression algorithm model is constructed in the following way: Preliminarily constructing a spectrum matrix based on the plurality of identification characteristic absorption peak data, wherein each row in the spectrum matrix represents absorbance data within a preset wavelength range; Constructing a concentration matrix in advance based on the sample concentrations of multiple metabolite markers, wherein each row in the concentration matrix corresponds to the concentration of one of the metabolite markers; Decomposing the spectrum matrix into latent variables, wherein the latent variables are linear combinations of the original spectra corresponding to the identification characteristic absorption peak data, and are used to explain the concentration change to the greatest extent; By minimizing the residual, a linear relationship between the latent variable and the concentration is established, and based on the leave-one-out method, the optimal number of latent variables is determined according to the linear relationship between the latent variable and the concentration, thereby obtaining a multivariate regression algorithm model.

7. The artificial intelligence-based pig immune function prediction system according to any one of claims 1 to 6, characterized in that: The key metabolic enzymes include one of the following: melatonin synthesis rate-limiting enzyme, methionine adenosine transferase; the immune effector molecules include at least one of the following: immunoglobulin A, immune interferon.

8. The artificial intelligence-based pig immune function prediction system according to any one of claims 1 to 6, characterized in that: The functional amino acid includes one of the following: γ-aminobutyric acid, serine, and tryptophan.

9. The artificial intelligence-based pig immune function prediction system according to any one of claims 1 to 6, characterized in that: The immune phenotype includes at least one of the following: blood immunoglobulin M content, immunoglobulin G content, interleukin-1β content, intestinal tissue secretory immunoglobulin A content, interleukin-22 content, CD68+ macrophage count, diarrhea rate, and morbidity.

10. A method for predicting pig immune function based on artificial intelligence, characterized in that: The prediction and display device according to any one of claims 1 to 9, wherein the method comprises: Based on the intestinal flora detection data, the relationship between the activity of key metabolic enzymes and the expression of immune effector molecules in pigs under stress is simulated based on a dynamic metabolic network to obtain metabolic immune interaction data, wherein the intestinal flora detection data is obtained by the detection device through detecting the intestinal flora corresponding to multiple omics of the pigs under stress; The metabolic immune interaction data is stored in a metabolic immune database, and based on the metabolic signature data and the metabolic immune interaction data stored in the metabolic immune database, a dose-effect relationship between functional amino acids and immune phenotypes is determined based on metabolic flux redirection to obtain metabolic spectrum characteristics of the pig under stress, wherein the metabolic immune database stores metabolic immune interaction data of multiple types of immune cells under multiple stress scenarios, and the metabolic signature data is obtained by detecting test result data of multiple metabolic markers of the pig under stress, and performing feature extraction based on the test result data; Based on the metabolic profile characteristics, amino acid ratios or dietary intervention strategies are predicted, generated, and displayed.

Images