Microbiome functional activity prediction method and system

Abstract

The invention relates to the technical field of microbial ecological information processing, and discloses a microbiome functional activity prediction method and system, and the method comprises the steps: collecting dynamic environment factor data in soil in real time through a sensor, and carrying out the preprocessing of the dynamic environment factor data; establishing a partial differential equation model based on the preprocessed data; the partial differential equation model comprises a diffusion term, a reaction term and an environmental factor coupling term; optimizing the differential equation model parameters by using a neural network, and carrying out transfer learning training on the partial differential equation model parameters based on historical data; and predicting the activity grade and function of the microorganisms on the basis of the optimized model for the functional activity of the microorganisms. The adaptability and generalization ability of the model in different soil areas are remarkably improved, and closed-loop processing from function activity numerical value prediction to function classification output is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of microbial ecological information processing, and particularly to a method and system for predicting the functional activity of a microbiome. Background Art

[0002] With the development of metagenomic sequencing technology and environmental omics, the soil microbiome, as an important biological component in the ecosystem, its functional activity has become an important entry point for studying agricultural sustainable development, soil health status, and ecological service functions. The current mainstream methods for microbiome functional analysis are mainly based on the annotation and abundance statistics of functional genes, often inferring the functional profile through 16S rRNA sequencing, or relying on metagenomic and metatranscriptomic data to construct functional pathway models to evaluate the potential functional capabilities of microbial communities. However, the above methods are mostly based on static data or assume that the microbial activity is relatively stable in a short period of time, ignoring the dynamic regulatory effect of environmental factors on microbial metabolic activity. With the continuous progress of high spatio-temporal resolution environmental sensors and multi-modal data acquisition means, existing studies have begun to explore the correlation modeling between environmental factors and microbial activity, but there are still significant challenges in the method mechanism. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by the present invention is: the technical problems of the existing methods for predicting the functional activity of a microbiome, such as the inability to dynamically reflect the influence of environmental factors, the lack of spatio-temporal modeling mechanism, and the insufficient prediction accuracy.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: A method for predicting the functional activity of a microbiome, comprising: collecting dynamic environmental factor data in soil in real time through sensors, and preprocessing the dynamic environmental factor data;

[0006] Based on the preprocessed data, a partial differential equation model is established; the partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term;

[0007] Using a neural network to optimize the parameters of the differential equation model, and performing transfer learning training on the parameters of the partial differential equation model based on historical data;

[0008] Based on the optimized model, predict the activity level and function of the microorganism for the microbial functional activity.

[0009] As a preferred scheme of the method for predicting the functional activity of a microbiome according to the present invention, wherein: the dynamic environmental factor data includes temperature, humidity, pH value, oxygen concentration, conductivity, dissolved oxygen, light intensity, soil water tension, and nutrients;

[0010] The preprocessing includes: cleaning the dynamic environmental factor data to remove noise and outliers, filling in missing data using spatial interpolation methods, and normalizing the data; integrating the data into unified environmental factor data using the weighted average method.

[0011] As a preferred embodiment of the method for predicting the functional activity of the microbiome according to the present invention, wherein: the diffusion term describes the diffusion process of the concentration of microbial metabolites in the soil, and is expressed by the formula:

[0012]

[0013] where C represents the concentration of microbial metabolites, D represents the diffusion coefficient, represents the Laplace operator in space;

[0014] The reaction term reflects the influence of microbial community activities on the concentration of their metabolites, and adjusts the production rate according to the change of environmental factor data, and is expressed by the formula:

[0015] R(C,E) = α·f(E)·C

[0016] where R(C,E) represents the reaction rate, E represents the environmental factor data, f(E) represents the influence function of the environmental factor data on the microbial metabolic reaction, α represents the reaction rate constant, and C represents the metabolite concentration;

[0017] The environmental factor coupling term is modeled through the coupling formula of environmental factors and microbial reaction rates to ensure the dynamic feedback of environmental changes on the functional activity of microorganisms, and is expressed by the formula:

[0018]

[0019] where λ represents the coupling coefficient, E

[0021] ,

[0020] ,

[0019] , , th , represents the threshold of environmental factors; β represents the adjustment coefficient.

[0020] As a preferred embodiment of the method for predicting the functional activity of the microbiome according to the present invention, wherein: optimizing the parameters of the differential equation model using a neural network includes: setting the diffusion coefficient, reaction rate coefficient, environmental coupling coefficient, steepness of the adjustment function, and environmental response threshold in the partial differential equation model as parameters to be optimized, and forming a parameter vector θ.

[0021] As a preferred embodiment of the method for predicting the functional activity of the microbiome according to the present invention, wherein: optimizing the parameters of the differential equation model using a neural network further includes: establishing a neural network model for predicting the parameters of the partial differential equation model based on historical soil environmental factor data and corresponding microbiome functional activity labels; integrating the preprocessed environmental factors into an input vector Xenv ;

[0022] The neural network model is a multi-layer perceptron architecture, including: an input layer, a first hidden layer, a second hidden layer, and an output layer;

[0023] The input layer receives X env as input; the first hidden layer maps X env to a high-dimensional space to obtain high-dimensional features h1; the second hidden layer extracts deep non-linear interaction information between environmental factors from the high-dimensional features to obtain h2; the output layer maps h2 to the target parameter dimension to obtain θ pred ;

[0024] Construct a joint loss function, substitute θ pred into the differential equation model to calculate the predicted microbial activity and the actual measured value to obtain the mean square error between them, and obtain the optimal parameter θ after convergence * .

[0025] As a preferred solution of the method for predicting the functional activity of the microbiome according to the present invention, wherein: the transfer learning training of the parameters of the partial differential equation model based on historical data includes: obtaining a historical soil environmental factor data set with microbial functional activity labels, and the historical data includes soil environmental factor sequences from multiple different regions or different time periods and corresponding microbial activity measurement values; using the historical data to train the neural network model to obtain the basic network structure and initial weight parameters;

[0026] In the case of limited target area environmental data samples, transfer the trained neural network model to the target area, keep the weights of the input layer and the first hidden layer unchanged, and only fine-tune the parameters of the remaining layer structures; based on the parameter vector output by the fine-tuned neural network construct a joint loss function, substitute it into the differential equation model to calculate the predicted microbial functional activity and the true measured value C true to obtain the mean square error between them, and obtain the optimal parameter after iterative training

[0027] As a preferred solution of the method for predicting the functional activity of the microbiome according to the present invention, wherein: predicting the activity level and function of the microbiome includes: substituting the optimized parameters θ of the partial differential equation model * into the differential equation model, and combining the real-time collected soil environmental factor data to calculate the predicted microbial functional activity value When , it is judged as high microbial activity; when , it is judged as medium microbial activity; when When it is [specific condition], it is determined as low microbial activity;

[0028] Based on the predicted activity level and the corresponding soil environmental factor data, the weighted logistic regression method is used to identify the functional types of the microbiome, with different functional labels as the classification targets, and the functional type with the highest corresponding probability is output; the functional types include, but are not limited to: nitrogen fixation, organic matter decomposition, and pathogen inhibition.

[0029] A microbiome functional activity prediction system, wherein:

[0030] The data module collects dynamic environmental factor data in the soil in real time through sensors and preprocesses the dynamic environmental factor data;

[0031] The partial differential equation model module establishes a partial differential equation model based on the preprocessed data; the partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term;

[0032] The parameter optimization module uses a neural network to optimize the parameters of the differential equation model and performs transfer learning training on the parameters of the partial differential equation model based on historical data;

[0033] The prediction module predicts the microbial functional activity based on the optimized model, and predicts the activity level and function of the microorganisms.

[0034] A computer device, including: a memory and a processor; the memory stores a computer program, and is characterized in that: when the processor executes the computer program, the steps of the method described in any one of the present inventions are implemented.

[0035] A computer-readable storage medium, on which a computer program is stored, and is characterized in that: when the computer program is executed by a processor, the steps of the method described in any one of the present inventions are implemented.

[0036] The beneficial effects of the present invention: The microbiome functional activity prediction method provided by the present invention can truly depict the spatio-temporal variation process of the microbial functional activity in the soil by introducing a coupled partial differential equation model and combining the dynamic perception and modeling of environmental factors; a neural network is used to construct a mapping relationship of differential equation parameters, and transfer learning training is performed based on historical environmental data, significantly improving the adaptability and generalization ability of the model in different soil regions; the model parameters are optimized by constructing a joint loss function, making the prediction results more accurate and stable; the microbial activity level is divided based on the interaction rule between the prediction result and the preset threshold, and further combined with the weighted logistic regression method to identify the specific functional types of the microbiome, realizing a closed-loop process from functional activity numerical prediction to functional classification output. Description of the Drawings

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0038] Figure 1 This is the overall flowchart of a method for predicting the functional activity of a microbiome provided in the first embodiment of the present invention. Detailed implementation manners

[0039] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific implementation manners of the present invention in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] Example 1, referring to Figure 1 , which is an embodiment of the present invention, provides a method for predicting the functional activity of a microbiome, including:

[0041] S1: Collect dynamic environmental factor data in the soil in real time through sensors, and preprocess the dynamic environmental factor data.

[0042] The dynamic environmental factor data includes temperature, humidity, pH value, oxygen concentration, conductivity, dissolved oxygen, light intensity, soil water tension, and nutrients.

[0043] The preprocessing includes: cleaning the dynamic environmental factor data, removing noise and outliers, filling in missing data using spatial interpolation methods, and normalizing the data. Using the weighted average method, the data is integrated into unified environmental factor data.

[0044] Furthermore, in the process of predicting the functional activity of the microbiome, environmental factors are the core driving variables that regulate the metabolic behavior and ecological response of microorganisms. The accuracy and timeliness of their data directly affect the reliability of the prediction model. In this step, by deploying multi-source soil environment sensors, key dynamic environmental factor data including temperature, humidity, pH value, oxygen concentration, conductivity, dissolved oxygen, light intensity, soil water tension, and nutrients are collected in real time, enabling the microbial activity modeling to move from static input to dynamic perception, laying a data foundation for subsequent dynamic modeling and feedback control. At the same time, this step supports flexible regulation of spatio-temporal resolution, enhancing the perception accuracy of the model for soil microenvironment changes.

[0045] Furthermore, data quality is improved through noise elimination and outlier identification, data continuity is enhanced by using spatial interpolation method to fill in missing points, the dimensionality impact between different dimensions is eliminated by standardization processing. Meanwhile, multi-channel sampling information is fused through a weighted average strategy to construct a unified and inputtable form of environmental factor vector, providing structured and high-quality input features for subsequent model calculation. This design significantly enhances the anti-interference ability and generalization ability of the model in non-ideal environments, providing an important guarantee for improving prediction accuracy and practicality.

[0046] S2: Based on the preprocessed data, a partial differential equation model is established. The partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term.

[0047] The diffusion term describes the diffusion process of the concentration of microbial metabolites in the soil, and is expressed by the formula:

[0048]

[0049] where C represents the concentration of microbial metabolites, D represents the diffusion coefficient, represents the Laplace operator in space.

[0050] The reaction term reflects the influence of microbial community activities on the concentration of their metabolites, and adjusts the generation rate according to the changes in environmental factor data, and is expressed by the formula:

[0051] R(C,E) = α·f(E)·C

[0052] where R(C,E) represents the reaction rate, E represents the environmental factor data, f(E) represents the influence function of environmental factor data on microbial metabolic reactions, α represents the reaction rate constant, and C represents the metabolite concentration.

[0053] The environmental factor coupling term is modeled through the coupling formula of environmental factors and microbial reaction rates to ensure the dynamic feedback of environmental changes on microbial functional activities, and is expressed by the formula:

[0054]

[0055] where λ represents the coupling coefficient, E th represents the threshold of environmental factors. β represents the adjustment coefficient.

[0056] Furthermore, during the prediction of microbiome functional activity, the metabolic behavior of microorganisms is affected by both time and space, and its response to external environmental factors is highly dynamic and non-linear. To accurately model such complex behaviors, the present invention establishes a partial differential equation model containing a diffusion term, a reaction term, and an environmental factor coupling term based on preprocessed data. Starting from physical mechanisms, this model uses the diffusion term to describe the spatial distribution dynamics of microbial metabolites in the soil, the reaction term to simulate the enhancement and inhibition processes of microbial community metabolic activities, and the coupling term to map the dynamic changes of various environmental factors into a feedback regulation mechanism of microbial functional activity, thereby constructing a continuous spatio-temporal prediction model that can reflect the dynamic response of microbial functions in real time.

[0057] Furthermore, by introducing a three-term coupling modeling mechanism, the partial differential equation model has good interpretability and generalizability. On the one hand, the model avoids the limitation of traditional methods that regard environmental factors as static inputs and improves the ability to depict the functional behaviors of microorganisms under dynamic environmental changes. On the other hand, the explicit modeling of spatial diffusion and environmental coupling enables the prediction model to not only output the numerical values of microbial functional activity intensities but also reflect their distribution and evolution trends at different time and space points. Compared with traditional statistical regression or black-box model methods, this modeling method has higher biological rationality and engineering practicability, laying a modeling foundation for achieving precise prediction and intelligent regulation for soil ecosystems.

[0058] S3: Optimize the parameters of the differential equation model using a neural network and perform transfer learning training on the parameters of the partial differential equation model based on historical data.

[0059] Optimizing the parameters of the differential equation model using a neural network includes: setting the diffusion coefficient, reaction rate coefficient, environmental coupling coefficient, steepness of the adjustment function, and environmental response threshold in the partial differential equation model as parameters to be optimized, forming a parameter vector θ.

[0060] Based on historical soil environmental factor data and corresponding microbial functional activity labels, establish a neural network model for predicting the parameters of the partial differential equation model. Integrate the preprocessed environmental factors into an input vector X env .

[0061] The neural network model is of a multi-layer perceptron architecture, including: an input layer, a first hidden layer, a second hidden layer, and an output layer.

[0062] The input layer receives X env as input.

[0063] The first hidden layer maps X env to a high-dimensional space to obtain high-dimensional features h1, and the formula is:

[0064] h1=ReLU(W1·X env +b1)

[0065] The second hidden layer extracts the deep nonlinear interaction information between environmental factors from the high-dimensional features to obtain h2, which is formulated as follows:

[0066] h2=ReLU(W2·h1+b2)

[0067] The output layer maps h2 to the target parameter dimension and obtains θ pred , the formula is:

[0068] θ pred =W3·h2+b3

[0069] Where h1 represents the output feature vector of the first hidden layer, ReLU(·) represents the activation function, W1 represents the weight matrix of the first hidden layer, X env represents the input vector, b1 represents the bias vector of the first hidden layer, h2 represents the output feature vector of the second hidden layer, W2 represents the weight matrix of the second hidden layer, b2 represents the bias vector of the second hidden layer, W3 represents the weight matrix of the output layer, and b3 represents the bias vector of the output layer.

[0070] Construct a joint loss function and transform θ pred Incorporating differential equation models into the calculation and prediction of microbial activity Compared with the actual measured value The mean square error between them is used to obtain the optimal parameter θ after convergence. * , the formula is:

[0071]

[0072] in, Indicates that θ pred The predicted concentrations were obtained after substituting into the PDE model. represents the actual observed microbial functional activity. η represents the regularization coefficient. : Indicates penalty control for parameter changes.

[0073]

[0074] in, Represents finding the parameter vector θ that minimizes the loss function L pred , the corresponding optimal parameter is expressed as θ*.

[0075] Obtain a historical soil environmental factor dataset with microbial functional activity tags, where the historical data includes soil environmental factor sequences from multiple different regions or different time periods and corresponding microbial activity measurement values. Use the historical data to train a neural network model to obtain the basic network structure and initial weight parameters.

[0076] In the case where the environmental data sample size of the target region is limited, transfer the trained neural network model to the target region, keeping the weights of the input layer and the first hidden layer unchanged, and only fine-tuning the parameters of the remaining layer structures. Based on the parameter vector output by the fine-tuned neural network Construct a joint loss function and substitute it into the differential equation model to calculate the predicted microbial functional activity and the mean square error with the true measurement value C true After iterative training, obtain the optimal parameters after convergence

[0077] Furthermore, by constructing a parameter prediction model based on a neural network, optimize the key control parameters in the partial differential equation, and introduce a transfer learning mechanism to improve the adaptability of the model in different soil regions. Compared with the traditional method of obtaining model parameters by empirical setting or global fitting, this step designs a multi-layer perceptron neural network structure that establishes a non-linear mapping relationship between environmental factor data and model parameters, which can automatically learn the influence mechanism of different environmental factors on core parameters such as diffusion coefficient, reaction rate, and coupling coefficient, and improve the scientificity and accuracy of parameter estimation. At the same time, by constructing a joint loss function, perform closed-loop correction on the parameter prediction results and the actual microbial functional activity to ensure the physical consistency and prediction effectiveness of the parameter optimization process.

[0078] Furthermore, introducing a transfer learning strategy can make full use of historical soil environmental data from multiple regions or different time periods to pre-train the neural network and transfer it to the target region with insufficient samples, reducing the dependence on large-scale labeled data, and improving the generalization ability and transfer adaptability of the model. By freezing the parameters of the input layer and some hidden layers and only fine-tuning the subsequent structures, not only can the source domain knowledge be effectively retained, but also the differences in the new regional environment can be quickly adapted, significantly shortening the training time and improving the prediction stability. This method realizes the automation, generalization, and high-precision feature extraction of parameter optimization, and can continuously output high-quality functional activity prediction results under dynamic environmental conditions, providing strong modeling support and intelligent means for soil microecological modeling.

[0079] S4: Based on the optimized model for microbial functional activity, predict the activity level and functions of microorganisms.

[0080] The parameters θ of the optimized partial differential equation model *Substitute into the differential equation model, combine with the soil environmental factor data collected in real time, and calculate the predicted microbial functional activity value When it is judged as high microbial activity. When it is judged as medium microbial activity. When it is judged as low microbial activity.

[0081] Based on the predicted activity level and the corresponding soil environmental factor data, a weighted logistic regression method is used to identify the functional types of the microbiome. Different functional labels are used as classification targets, and the functional type with the highest corresponding probability is output. The functional types include, but are not limited to: nitrogen fixation, organic matter decomposition, and pathogen inhibition.

[0082] Furthermore, by substituting the optimized partial differential equation model parameters into the complete prediction model, combining with the environmental factor data collected in real time, dynamically calculating the microbial functional activity value, and based on the set activity level threshold, using clear interaction rules for activity level classification, the quantitative and hierarchical expression of the microbial functional state is realized. Compared with the traditional method that only relies on gene abundance or empirical indicators for function inference, this method significantly improves the scientificity and adaptability of activity assessment, and is especially suitable for the actual soil environment with frequent ecological changes or continuous data collection.

[0083] Even further, a weighted logistic regression method is introduced to conduct a fusion analysis of the predicted microbial activity level and environmental factor characteristics, so as to realize the intelligent identification and probability output of microbial functional types (such as nitrogen fixation, organic matter decomposition, pathogen inhibition, etc.). This method not only has good model interpretability, but also can be extended to a multi-functional identification model to adapt to different task requirements. By establishing a quantitative mapping relationship between the functional level and the specific ecological role, the present invention is superior to the prior art in terms of functional prediction accuracy, functional type discrimination sensitivity, and ecological adaptation ability, and has significant technical promotion value.

[0084] Embodiment 2, an embodiment of the present invention, provides a microbial group functional activity prediction system, including:

[0085] A data module that collects dynamic environmental factor data in the soil in real time through sensors and preprocesses the dynamic environmental factor data.

[0086] A partial differential equation model module that establishes a partial differential equation model based on the preprocessed data. The partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term.

[0087] A parameter optimization module that uses a neural network to optimize the differential equation model parameters and conducts transfer learning training on the partial differential equation model parameters based on historical data.

[0088] A prediction module, based on the optimized model, predicts the activity level and functions of microorganisms with respect to the functional activity of microorganisms.

[0089] Example 3, an embodiment of the present invention, which is different from the previous two embodiments in that:

[0090] If the said function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, etc., all kinds of media that can store program codes.

[0091] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0092] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), optical fiber devices, and portable compact disc read-only memories (CDROMs). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or, if necessary, other suitable processing, and then stored in a computer memory.

[0093] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0094] Example 4 is an embodiment of the present invention, which provides a method and system for predicting the functional activity of a microbiome. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0095] Seven groups of soil samples (Samples A to G) were selected for the experiment, and their corresponding environmental parameters (including temperature, humidity, pH value, dissolved oxygen, conductivity, etc.) were simulated. A preprocessing module was introduced to standardize and denoise the data. At the same time, the data was input into the trained neural network model to predict the key parameters (such as diffusion coefficient, reaction rate coefficient, coupling coefficient, etc.) required by the partial differential equation model, and the predicted value of the microbial functional activity of each group of samples under the current environment was output by combining model solving. In the experiment, the high activity threshold was set as θ1 = 0.7, and the low activity threshold was set as θ2 = 0.4. When the predicted value it was determined as high activity, determined as medium activity, determined as low activity. Based on the determination of the activity level, the predicted level and the environmental factor vector combination were further input into the weighted logistic regression classifier to achieve probability output and optimal classification of the microbial functional types (nitrogen fixation, organic matter decomposition, pathogen inhibition). To verify the prediction accuracy, the "true activity value" of each group of samples was artificially simulated as a control, the prediction deviation was calculated, and the prediction ability and stability of the model were analyzed accordingly. This experiment was based on the actual modeling process, covering the entire process of data preprocessing, parameter optimization, prediction calculation, and function judgment, ensuring the integrity and verifiability of the experimental process.

[0096] In this embodiment, seven groups of samples under different soil environments (named Samples A to G respectively) were selected, their key environmental factors were simulated and assigned values and standardized, and the microbial functional activity of each group of samples was predicted and functionally identified by combining the method of the present invention.

[0097] Specifically, the temperature of sample A is 26.7°C, the humidity is 72.4%, the pH value is 6.9, the dissolved oxygen is 9.3 mg / L, and the conductivity is 390.5 μS / cm. Under these conditions, the predicted functional activity is 0.83, and the corresponding actual activity is 0.79, which is judged as high activity, and the function is identified as "nitrogen fixation";

[0098] The environmental conditions of sample B are 24.2°C, 65.1%, pH value 7.2, dissolved oxygen 3.5 mg / L, and conductivity 607.3 μS / cm. The predicted activity is 0.59, and the actual activity is 0.56, which is judged as medium activity, and the function is identified as "organic matter decomposition";

[0099] When the environmental factors of sample C are 27.1°C, 60.8%, pH value 7.6, dissolved oxygen 8.2 mg / L, and conductivity 402.2 μS / cm, the predicted activity reaches 0.77, and the actual value is 0.75, which is judged as high activity, and the function type is "pathogen inhibition";

[0100] The environment of sample D is 22.5°C, 51.7%, pH 6.1, dissolved oxygen 4.1 mg / L, and conductivity 735.8 μS / cm. The predicted activity is 0.48, and the actual activity is 0.45, which is judged as medium activity, and the function is identified as "nitrogen fixation";

[0101] When sample E is under relatively low temperature and humidity conditions (20.8°C, 53.6%, pH value 5.7, dissolved oxygen 5.4 mg / L, and conductivity 282.9 μS / cm), the predicted activity is 0.43, and the actual activity is 0.41, which belongs to medium activity, and the function is "pathogen inhibition";

[0102] Sample F is in the condition of extremely low dissolved oxygen and acidic soil (temperature 21.9°C, humidity 69.4%, pH 5.5, dissolved oxygen 2.8 mg / L). The predicted activity is only 0.29, and the actual value is 0.26, which is judged as low activity, and the function is "organic matter decomposition";

[0103] When sample G is under the conditions of 25.5°C, 78.3%, pH 7.0, dissolved oxygen 7.9 mg / L, and conductivity 518.3 μS / cm, the predicted activity is 0.81, and the actual activity is 0.78, which is high activity, and the function is identified as "pathogen inhibition".

[0104] The deviation between the predicted result and the actual activity value is controlled within ±0.07. The predicted grade and function output also show good discrimination in ecological significance, which supports the effectiveness of the method of the present invention in modeling the microbial functional state in complex soil environments.

[0105] As can be seen from the data, the predicted microbial functional activity values of different soil samples under multi-environmental factor combinations show obvious numerical differences. Taking sample A and sample F as examples, when their temperature and pH conditions are similar, due to the significant difference in dissolved oxygen (9.3 mg / L for sample A and only 2.8 mg / L for sample F), the predicted activity values are 0.83 and 0.29 respectively, successfully reflecting the sensitive response ability of the model of the present invention to key environmental regulation factors. Further, by comparing with the "actual activity value", the average deviation rate is controlled within ±0.07, indicating that this method has good prediction accuracy and can stably capture the dynamic impact of environmental changes on microbial activity.

[0106] In terms of function recognition, the model combines the activity level and environmental factors to accurately judge the function type of the sample through weighted logistic regression. For example, both sample C and sample G in the prediction results are "high activity", but are respectively identified as "nitrogen fixation" and "pathogen inhibition" in the function output, reflecting that this model can not only quantify the degree of microbial activity, but also refine its functional role, realizing a deeper ecological function classification. Compared with the traditional method of inferring functions relying on gene abundance, the method of the present invention establishes a ternary mapping relationship of "functional activity - environmental factor - ecological function", with multiple advantages of dynamics, accuracy and interpretability.

[0107] In summary, as can be seen from the experimental data of the embodiments, the method of the present invention shows better technical performance than the existing static analysis methods in aspects such as dynamic modeling, parameter generalization, activity determination and function recognition, effectively overcoming the deficiencies of traditional models being insensitive to spatio-temporal dynamic responses and having rough function classification, and having significant academic innovation value and practical application prospects.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the functional activity of a microbiome, characterized in that, Including: Collecting dynamic environmental factor data in soil in real time through sensors, and preprocessing the dynamic environmental factor data; Based on the preprocessed data, establishing a partial differential equation model; The partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term; Using a neural network to optimize the parameters of the differential equation model, and performing transfer learning training on the parameters of the partial differential equation model based on historical data; Based on the optimized model, predicting the microbial functional activity, the activity level and function of the microorganisms.

2. The method for predicting the functional activity of a microbiome according to claim 1, wherein: The dynamic environmental factor data includes temperature, humidity, pH value, oxygen concentration, conductivity, dissolved oxygen, light intensity, soil water tension, and nutrients; The preprocessing includes: cleaning the dynamic environmental factor data, removing noise and outliers, filling in missing data using spatial interpolation methods, and normalizing the data; using the weighted average method to integrate the data into unified environmental factor data.

3. The method for predicting the functional activity of a microbiome according to claim 2, wherein: The diffusion term describes the diffusion process of the concentration of microbial metabolites in soil, and is expressed by the formula: Where C represents the concentration of microbial metabolites and D represents the diffusion coefficient, represents the Laplace operator in space; The reaction term reflects the influence of microbial community activities on the concentration of its metabolites, and adjusts the generation rate according to the changes in environmental factor data, and is expressed by the formula: R(C,E)=α·f(E)·C Wherein, R(C,E) represents the reaction rate, E represents the environmental factor data, f(E) represents the influence function of the environmental factor data on the microbial metabolic reaction, α represents the reaction rate constant, and C represents the metabolite concentration; The environmental factor coupling term is modeled through the coupling formula of the environmental factor and the microbial reaction rate to ensure the dynamic feedback of environmental changes on the microbial functional activity, and is expressed by the formula: Among them, λ represents the coupling coefficient, and E th represents the threshold value of the environmental factor; β represents the adjustment coefficient.

4. The method for predicting the functional activity of a microbiome according to claim 3, wherein: Using a neural network to optimize the parameters of the differential equation model includes: setting the diffusion coefficient, reaction rate coefficient, environmental coupling coefficient, steepness of the adjustment function, and environmental response threshold in the partial differential equation model as parameters to be optimized, and forming a parameter vector θ.

5. The method for predicting the functional activity of a microbiome according to claim 4, wherein: Optimizing the parameters of the differential equation model using a neural network further includes: establishing a neural network model for predicting the parameters of the partial differential equation model based on historical soil environmental factor data and corresponding microbial functional activity labels; integrating the preprocessed environmental factors into an input vector X env ; The neural network model is a multi-layer perceptron architecture, including: an input layer, a first hidden layer, a second hidden layer, and an output layer; The input layer receives X env as input; the first hidden layer maps X env to a high-dimensional space to obtain the high-level feature h1; the second hidden layer extracts the deep non-linear interaction information between environmental factors from the high-dimensional features to obtain h2; the output layer maps h2 to the target parameter dimension to obtain θ pred ; Construct a combined loss function and substitute θ pred into the differential equation model to calculate the predicted microbial activity and the mean square error between the predicted microbial activity and the actual measured value to obtain the optimal parameter θ after convergence * .

6. The method for predicting the functional activity of the microbiome according to claim 5, wherein: Performing transfer learning training on the parameters of the partial differential equation model based on historical data includes: obtaining a historical soil environmental factor data set with microbial functional activity labels, where the historical data includes soil environmental factor sequences and corresponding microbial activity measurement values from multiple different regions or different time periods; using the historical data to train the neural network model to obtain the basic network structure and initial weight parameters; In the case of limited environmental data sample size in the target area, the trained neural network model is migrated to the target area while keeping the weights of the input layer and the first hidden layer unchanged, and only the parameters of the remaining layer structures are fine-tuned; based on the parameter vector output by the fine-tuned neural network Construct a joint loss function and substitute it into the differential equation model to calculate the predicted microbial functional activity and the mean square error with the true measured value C true After iterative training, the optimal converged parameters are obtained 7. The method for predicting the functional activity of the microbiome according to claim 6, wherein: Predicting the activity level and functions of the microbiome includes: substituting the optimized partial differential equation model parameter θ * into the differential equation model, and combining with the real-time collected soil environmental factor data to calculate the predicted microbial functional activity value When it is judged as high microbial activity; when it is judged as medium microbial activity; when it is judged as low microbial activity; Based on the predicted activity level and the corresponding soil environmental factor data, using the weighted logistic regression method to identify the functional types of the microbiome, taking different functional labels as classification targets, and outputting the functional type with the highest corresponding probability; the functional types include but are not limited to: nitrogen fixation, decomposition of organic matter, and pathogen inhibition.

8. A microbial group functional activity prediction system adopting the method according to any one of claims 1-7, characterized in that: A data module that collects dynamic environmental factor data in soil in real time through sensors, and preprocesses the dynamic environmental factor data; The partial differential equation model module establishes a partial differential equation model based on the preprocessed data; The partial differential equation model includes: a diffusion term, a reaction term, and an environmental factor coupling term; The parameter optimization module uses a neural network to optimize the parameters of the differential equation model and performs transfer learning training on the parameters of the partial differential equation model based on historical data; The prediction module predicts the activity level and function of microorganisms based on the optimized model for the microbial functional activity.

Images