Method and device for predicting classification of new pollutants

Abstract

The application provides a new pollutant classification prediction method and related device, and relates to the technical field of new pollutant classification. Including the following steps: collecting the chemical information (molecular structure, molecular fingerprint, physicochemical property) of the target pollutant and performing standardization processing; based on Tanimoto similarity, preliminary classification is calculated; the migration path and transformation behavior of the pollutant in the environment are simulated by using a graph neural network; the multi-mechanism comprehensive evaluation is comprehensively considered in terms of persistence, migration and biological accumulation; an environmental factor correction model is introduced, and the score is corrected according to parameters such as temperature, pH value, dissolved oxygen concentration and the like; the molecular structure, migration path, mechanism characteristics and environmental correction results are integrated to calculate the final classification prediction value; and the risk level (high, medium and low risk) is divided according to the prediction value. Through machine learning technology, the pollutant risk is comprehensively evaluated, the evaluation accuracy and reliability are improved, and a scientific basis is provided for pollution prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new pollutant classification, in particular to a new pollutant classification prediction method and related device. BACKGROUND

[0002] With the acceleration of industrialization, emerging contaminants (ECs) such as pharmaceutical residues and microplastics pose a serious challenge to ecological safety due to their complex structure, variable environmental behavior and multiple harmful mechanisms. Traditional classification methods rely on molecular fingerprint matching or static physicochemical indicators, which are difficult to dynamically quantify the migration and transformation process of pollutants in the environment, and ignore the dynamic influence of multi-mechanism synergy and environmental factors. There are two major limitations in the existing technology: first, the data model breaks the link between the inherent properties of pollutants and environmental parameters, and second, there is a lack of comprehensive evaluation system for persistence, bioaccumulation and other core mechanisms.

[0003] Therefore, how to improve the classification accuracy of new pollutants has become a technical problem to be solved. SUMMARY

[0004] In order to improve the classification accuracy of new pollutants, the present application provides a new pollutant classification prediction method and related device.

[0005] In a first aspect, the present application provides a new pollutant classification prediction method using the following technical solution:

[0006] A new pollutant classification prediction method comprises:

[0007] S1: Collecting chemical information of a target pollutant, the chemical information including molecular structure description, molecular fingerprint and physicochemical properties, and performing standardization processing on the chemical information;

[0008] S2: Calculating the chemical information after standardization processing based on a Tanimoto similarity calculation method to obtain a calculation result, and combining the calculation result with existing classification data to perform weighted integration to preliminarily predict the classification of the target pollutant and generate a preliminary classification evaluation result;

[0009] S3: Simulating the migration path of the target pollutant in the environment using a graph neural network and analyzing the migration, transformation and degradation behavior of the target pollutant in different environmental media to generate an analysis result;

[0010] S4: According to the analysis result, considering the persistence, mobility and bioaccumulation of the target pollutant, performing multi-mechanism comprehensive evaluation through weighted summation to calculate the comprehensive mechanism characteristic score corresponding to the target pollutant;

[0011] S5: Introducing an environmental factor correction model, according to temperature, pH value, dissolved oxygen concentration, humidity and light intensity, correcting the intrinsic characteristic score of the target pollutant;

[0012] S6: Based on the preliminary classification evaluation result, the analysis result, the intrinsic characteristic score and the environmental correction result, calculating the final classification prediction value of the target pollutant;

[0013] S7: According to the final classification prediction value, dividing the risk level of the target pollutant, and outputting the classification result.

[0014] Optionally, the step of collecting the chemical information of the target pollutant, the chemical information including: molecular structure description, molecular fingerprint and physicochemical property, and standardizing the chemical information, comprises:

[0015] Collecting the chemical information of the target pollutant , representing the molecular structure description in the chemical information as a set composed of a series of atoms:

[0016]

[0017] Each atom has its unique chemical property ;

[0018] Converting the molecule into a fixed-dimensional vector by using MACCS or Morgan fingerprint, assuming the molecular fingerprint of the target pollutant is a vector

[0019]

[0020] Wherein, each component represents whether there is a certain structure or functional group;

[0021] After standardization processing, the physicochemical property is represented as:

[0022]

[0023] Wherein and are the mean and standard deviation of the feature respectively.

[0024] Optionally, the Tanimoto similarity calculation formula in the step S2 is:

[0025]

[0026] Wherein, the molecular fingerprint of the target pollutant is , and the molecular fingerprint of the i-th pollutant in the database is The molecular fingerprint of a known classified compound is , inner product: , vector norm: .

[0027] Optionally, the migration path analysis is implemented through a graph neural network, specifically including:

[0028] The environment is divided into multiple nodes, each node representing a specific environmental medium, and the characteristics of each node include the migration rate, adsorption capacity and degradation rate of pollutants;

[0029] The migration path of the pollutant is modeled using a graph neural network, and the input of the network includes the physical and chemical properties of the pollutant and environmental factors to calculate the migration probability between different environmental nodes, and the output is the migration path and expected residence time of the pollutant in the environment.

[0030] Optionally, the comprehensive mechanism characteristic score is calculated by the following formula:

[0031]

[0032] Wherein, , , respectively represent the weights of the persistence, migration and bioaccumulation mechanisms, represent the pollutant, , and represent the scores of the pollutant in terms of persistence, migration and bioaccumulation.

[0033] Optionally, in the step S5, the environmental correction model is calculated by the following formula:

[0034]

[0035] Wherein, is the environmental correction factor, and the intrinsic characteristics of the pollutant are scored as ;

[0036]

[0037] The environmental parameters are denoted as , is the score of the inherent properties of the pollutant.

[0038] Optionally, the final classification prediction value is calculated by the following formula:

[0039]

[0040] Wherein represents the preliminary classification evaluation result based on the molecular structure, : represents the analysis result considering the migration path and transformation behavior, : represents the intrinsic property score evaluated based on multiple classification mechanisms, : represents the classification property after environmental correction;

[0041] weighting parameters all satisfy:

[0042] 。

[0043] In a second aspect, the application provides a new pollutant classification prediction system, comprising:

[0044] an information collection module for collecting chemical information of a target pollutant, the chemical information including molecular structure description, molecular fingerprint, and physicochemical properties, and performing standardization processing on the chemical information;

[0045] an information collection module for collecting chemical information of a target pollutant, the chemical information including molecular structure description, molecular fingerprint, and physicochemical properties, and performing standardization processing on the chemical information;

[0046] a preliminary classification evaluation result module for calculating the standardization-processed chemical information based on a Tanimoto similarity calculation method to obtain a calculation result, and combining the calculation result with existing classification data to preliminarily predict the classification of the target pollutant and generate a preliminary classification evaluation result;

[0047] an analysis result module for simulating the migration path of the target pollutant in the environment by using a graph neural network and analyzing the migration, transformation, and degradation behavior of the target pollutant in different environmental media to generate an analysis result;

[0048] a comprehensive mechanism feature score module for comprehensively considering the persistence, mobility, and bioaccumulation of the target pollutant according to the analysis result, and performing multi-mechanism comprehensive evaluation by weighted summation to calculate the comprehensive mechanism feature score corresponding to the target pollutant;

[0049] an intrinsic property score module for introducing an environmental factor correction model, and correcting the intrinsic property score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity, and light intensity;

[0050] a final classification prediction value module for calculating the final classification prediction value of the target pollutant based on the preliminary classification evaluation result, the analysis result, the intrinsic property score, and the environmental correction result;

[0051] an output module for dividing the risk level of the target pollutant according to the final classification prediction value, and outputting a classification result.

[0052] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the processor executes the method as described above when running computer instructions stored in the memory.

[0053] In a fourth aspect, the present application provides a computer readable storage medium comprising instructions which, when executed on a computer, cause the computer to perform the method as described above.

[0054] In summary, the present application includes the following beneficial technical effects:

[0055] The present application collects the chemical information (molecular structure, molecular fingerprint, physicochemical properties) of the target pollutants and performs standardization processing; preliminary classification is based on Tanimoto similarity calculation; the migration path and transformation behavior of pollutants in the environment are simulated by using graph neural network; multi-mechanism comprehensive evaluation is comprehensively considered in terms of persistence, mobility and bioaccumulation; an environmental factor correction model is introduced to correct the score according to parameters such as temperature, pH value, dissolved oxygen concentration, etc.; the final classification prediction value is calculated by integrating the results of molecular structure, migration path, mechanism characteristics and environmental correction; and the risk level (high, medium and low risk) is divided according to the prediction value. Through machine learning technology, the pollutant risk is comprehensively evaluated, the evaluation accuracy and reliability are improved, and a scientific basis is provided for pollution prevention and control. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a computer device structure schematic diagram of a hardware running environment related to an embodiment scheme of the present application;

[0057] Figure 2 is a flowchart of a first embodiment of the new pollutant classification prediction method of the present application;

[0058] Figure 3 is a structure block diagram of a first embodiment of the new pollutant classification prediction system of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described in detail below by referring to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0060] Referring to Figure 1 , Figure 1 is a computer device structure schematic diagram of a hardware running environment related to an embodiment scheme of the present application.

[0061] As Figure 1As shown, the computer device can include a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is configured to realize the connection and communication between the components. The user interface 1003 can include a display, an input unit such as a keyboard, and can further include a standard wired interface, a wireless interface. The network interface 1004 can optionally include a standard wired interface, a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 can be a high-speed random access memory (RAM), and can also be a stable non-volatile memory (NVM), such as a disk memory. The memory 1005 can also be a storage device independent of the aforementioned processor 1001.

[0062] Those skilled in the art can understand that the structure shown in the above Figure 1 The structure shown in the above does not constitute a limitation on the computer device, and can include more or fewer components than those shown, or combine certain components, or different component arrangements.

[0063] As shown in the above Figure 1 The memory 1005 as a storage medium can include an operating system, a network communication module, a user interface module, and a new pollutant classification prediction program.

[0064] In the computer device shown in the above Figure 1 In the computer device shown in the above, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the present application can be arranged in the computer device, and the computer device calls the new pollutant classification prediction program stored in the memory 1005 through the processor 1001, and executes the new pollutant classification prediction method provided in the embodiments of the present application.

[0065] The embodiments of the present application provide a new pollutant classification prediction method, which refers to Figure 2 , Figure 2 The flowchart of the first embodiment of the new pollutant classification prediction method of the present application is shown in the above.

[0066] In the present embodiment, the new pollutant classification prediction method includes the following steps:

[0067] S1: Collect the chemical information of the target pollutant, including molecular structure description, molecular fingerprint, and physicochemical properties, and standardize the chemical information.

[0068] It can be understood that traditional new pollutant classification and prediction methods, such as QSAR models and molecular docking techniques, can provide some characteristic information to some extent, but face the following challenges:

[0069] Data scarcity: For new pollutants, there is a lack of experimental data and ready-made classification models.

[0070] High cost and low efficiency: Existing classification experiments and simulation methods require a large amount of experimental data support, and are usually high in cost and long in experimental period.

[0071] Poor comprehensiveness: Existing technologies cannot effectively combine multi-dimensional data of pollutants (such as chemical structure, environmental behavior, migration path, etc.) for comprehensive analysis.

[0072] Insufficient environmental behavior prediction: Existing methods are difficult to fully consider the migration, transformation and accumulation processes of pollutants under different environmental conditions.

[0073] It should be noted that the step of collecting the chemical information of the target pollutant, including molecular structure description, molecular fingerprint, and physicochemical properties, and standardizing the chemical information, includes: Collecting the chemical information of the target pollutant

[0074]

[0075] Each atom has its unique chemical properties ;

[0076] Convert the molecule into a fixed-dimensional vector using MACCS or Morgan fingerprint, set the molecular fingerprint of the molecule as a vector;

[0077]

[0078] where each component indicates whether a certain structure or functional group exists;

[0079] After standardization, the physicochemical properties are represented as :

[0080]

[0081] where and are the mean and standard deviation of the feature, respectively.

[0082] S2: Calculate the normalized chemical information based on the Tanimoto similarity calculation method to obtain a calculation result, and combine the calculation result with the existing classification data to obtain a preliminary classification prediction of the target pollutant and generate a preliminary classification evaluation result.

[0083] It can be understood that the Tanimoto similarity calculation formula in step S2 is:

[0084]

[0085] wherein the molecular fingerprint of the target pollutant is , the molecular fingerprint of the i-th known classified compound in the database is , the inner product is , and the vector norm is . .

[0086] It should be noted that the existing classification database is used in combination with molecular structure similarity to make a preliminary classification prediction. The key here is to use a data-driven method to infer the potential class of the target pollutant by comparing the classification labels of similar structured compounds.

[0087] In specific implementation, the step of weighted integration of classification data includes: assuming that there are compounds similar in structure to the target compound in the database, each compound has a known classification label (such as persistent organic pollutants, endocrine disruptors, etc.), and the preliminary prediction class of the target molecule can be obtained by weighted voting:

[0088]

[0089] wherein I(·) is an indicator function, and is the set of all possible classification labels.

[0090] S3: Simulate the migration path of the target pollutant in the environment using a graph neural network and analyze the migration, transformation and degradation behavior of the target pollutant in different environmental media to generate an analysis result.

[0091] ​It should be noted that in step S3, the migration path analysis is implemented through a graph neural network, specifically including: dividing the environment into multiple nodes, each node representing a specific environmental medium, the characteristics of each node including the migration rate of pollutants, adsorption capacity, degradation rate; using a graph neural network to model the migration path of pollutants, the input of the network includes the physical and chemical properties of the pollutants and environmental factors, to calculate the migration probability between different environmental nodes, output the migration path and expected residence time of the pollutants in the environment.

[0092] In a specific implementation, the step of evaluating environmental behavior and migration path includes:

[0093] 1. Migration path analysis

[0094] The migration path of pollutants in the environment is mainly affected by their physical and chemical properties (such as solubility, volatility, hydrophilicity, etc.) and environmental conditions (such as temperature, pH, wind speed, etc.). In order to accurately predict the migration path of pollutants in different environmental media (such as water, soil, atmosphere, etc.), this embodiment uses a graph neural network (Graph Neural Network, GNN) to simulate. The specific process is as follows:

[0095] 1.1 Environmental medium modeling: The environment is divided into multiple nodes, each node representing a specific environmental medium (for example, water body, soil, atmosphere, etc.). The characteristics of each node include the migration rate of pollutants, adsorption capacity, degradation rate, etc.

[0096] 1.2 Migration path calculation: Use a graph neural network to model the migration path of pollutants, the input of the network includes the physical and chemical properties of the pollutants and environmental factors (such as temperature, humidity, pH value). Through GNN, the migration probability of pollutants between each environmental node is calculated, and the migration path of pollutants in the environment is obtained.

[0097] 1.3 Migration path output: The model will output the migration probability distribution of pollutants in different environmental media, as well as the expected residence time and migration range of pollutants in each environmental medium.

[0098] 2. Transformation behavior modeling:

[0099] Pollutants may undergo transformation in the environment, such as photodegradation, biodegradation, or chemical reaction, etc. This embodiment defines a transformation function to describe the transformation behavior of pollutants, and uses the following steps for modeling;

[0100]

[0100] 2.1 Transformation mechanism definition: The transformation behavior of pollutants is determined by different transformation mechanisms, such as chemical degradation, photodegradation, microbial degradation, etc. The transformation rate of each mechanism can be estimated through experimental data or literature.

[0101] 2.2 Transformation product prediction: Contaminants It is possible that a contaminant can be transformed into multiple products in the environment. Each transformation mechanism will produce a certain set of transformation products , which can have an impact on the final classification result of the contaminant.

[0102] 2.3 Transformation impact assessment: By establishing a dynamic model of contaminant transformation, combined with environmental conditions and initial properties of the contaminant, the changes in classification characteristics of the contaminant due to transformation are assessed. For example, some contaminants may become more difficult to bioaccumulate after biodegradation, while some transformation products may have higher toxicity.

[0103] S4: According to the analysis results, the persistence, mobility and bioaccumulation of the target contaminant are comprehensively considered, and a multi-mechanism comprehensive evaluation is carried out by weighted summation to calculate the comprehensive mechanism characteristic score corresponding to the target contaminant.

[0104] It can be understood that in the step S4, the comprehensive mechanism characteristic score is calculated by the following formula:

[0105]

[0106] wherein, , , represent the weights of the persistence, mobility and bioaccumulation mechanisms respectively, represents the contaminant, , and represent the scores of the contaminant in terms of persistence, mobility and bioaccumulation respectively.

[0107] In specific implementation, mechanism-related classification modeling is carried out;

[0108] In the classification process, each mechanism has different effects on the performance of the contaminant. We consider the following main mechanisms:

[0109] Persistence: represents the time a contaminant exists in the environment, usually represented by half-life or degradation rate. The persistence of a contaminant is closely related to its chemical stability and environmental conditions (such as temperature, light, etc.).

[0110] Mobility: describes the migration ability of a contaminant in different media. The mobility of a contaminant is affected by its physicochemical properties (such as partition coefficient, solubility, etc.), and a contaminant with strong mobility is more likely to spread in the environment.

[0111] Bioaccumulation: refers to the ability of a pollutant to accumulate in living organisms, usually measured by the bioaccumulation factor (BAF). Pollutants with high bioaccumulation may have long-term effects on ecosystems.

[0112] In specific implementations, the step of weight setting and optimization includes setting weights for mechanisms such as persistence, mobility, and bioaccumulation, which can be optimized through machine learning methods so that these weights can better adapt to the actual experimental data or simulated environment of pollutants. The optimization goal is to minimize the error between the comprehensive score and the actual observation value. Set the comprehensive mechanism score of the target pollutant as , which is defined as:

[0113]

[0114] where , , are the weights of persistence, mobility, and bioaccumulation, respectively, , and are the scores of the pollutant under these mechanisms.

[0115] The goal of optimization is to adjust these weights so that the prediction result is as close as possible to the actual classification label or experimental data of the pollutant. The specific objective function is:

[0116]

[0117] where represents the true classification score (experimental observation data or expert label) of the th pollutant, is the number of samples.

[0118] S5: Introduce an environmental factor correction model to correct the intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity, and light intensity.

[0119] It should be noted that in the step S5, the environmental correction model is calculated by the following formula:

[0120]

[0121] where is the environmental correction factor, and the intrinsic characteristic of the pollutant is scored as ;

[0122]

[0123] The environmental parameters are denoted as , Score of intrinsic property of the pollutant.

[0124] S6: Calculate the final classification prediction value of the target pollutant based on the preliminary classification evaluation result, the analysis result, the intrinsic property score and the environmental correction result.

[0125] It can be understood that the final classification prediction value is calculated by the following formula:

[0126]

[0127] Wherein represents the preliminary classification evaluation result based on the molecular structure, : represents the analysis result considering the migration path and transformation behavior, represents the intrinsic property score evaluated based on multiple classification mechanisms, represents the classification characteristics after environmental correction;

[0128] The weight parameters all satisfy:

[0129] 。

[0130] S7: According to the final classification prediction value, the risk level of the target pollutant is divided, and the classification result is output.

[0131] In specific implementation, according to the value of , the classification level of the pollutant can be divided:

[0132]

[0133] Wherein and are threshold values determined according to experimental data or industry standards.

[0134] Similarity calculation and preliminary classification: based on the Tanimoto similarity calculation method, the standardized chemical information is calculated, combined with the existing classification data for weighted integration, and the classification of the target pollutant is preliminarily predicted.

[0135] Migration path simulation: using graph neural network model to simulate the migration path of target pollutant in the environment, and analyzing its migration, transformation and degradation behavior in different environmental media.

[0136] Multi-mechanism comprehensive evaluation: considering the persistence, mobility and bioaccumulation of the target pollutant, the multi-mechanism comprehensive evaluation is carried out by weighted summation, and the comprehensive mechanism characteristic score is calculated.

[0137] Environmental factor correction: Introduce an environmental factor correction model to correct the intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity, and light intensity.

[0138] Final classification prediction: Integrate the molecular structure, migration path, mechanism characteristics, and environmental correction results to calculate the final classification prediction value of the target pollutant.

[0139] Risk level division: According to the final classification prediction value, divide the risk level of the target pollutant, and output the classification result. The risk level is divided into three levels: high risk, medium risk, and low risk.

[0140] In this embodiment, the chemical information (molecular structure, molecular fingerprint, and physicochemical properties) of the target pollutant is collected and standardized; a preliminary classification is made based on Tanimoto similarity; a graph neural network is used to simulate the migration path and transformation behavior of the pollutant in the environment; a multi-mechanism comprehensive evaluation is made considering persistence, mobility, and bioaccumulation; an environmental factor correction model is introduced to correct the score according to parameters such as temperature, pH value, and dissolved oxygen concentration; the final classification prediction value is calculated by integrating the molecular structure, migration path, mechanism characteristics, and environmental correction results; and the risk level (high, medium, and low risk) is divided according to the prediction value. Through machine learning technology, the pollutant risk is comprehensively evaluated, improving the evaluation accuracy and reliability, and providing a scientific basis for pollution prevention.

[0141] The content of the second embodiment of the present application includes:

[0142] Step 1, collect the chemical information of the target pollutant, which includes molecular structure description, molecular fingerprint, and physicochemical properties, and standardize the chemical information.

[0143] Specifically, the molecular structure data of the target pollutant is input into a standardization processing module for standardization processing, and the standardization processing method can use Z-Score standardization or Min-Max standardization method. The attention mechanism is used to find the substructure that contributes to the prediction attribute value, and the attention mechanism can use multi-head self-attention mechanism. The problem of insufficient sample size is solved by parameter transfer, and the parameter transfer method can use domain adversarial-based parameter transfer method. An interspecies extrapolation normalization model is established to increase the number of species in the derivation of soil environmental benchmarks, making the derived benchmarks more scientific and reliable, thereby protecting more species. The interspecies extrapolation normalization model can use a machine learning-based interspecies extrapolation model.

[0144] Step 2, calculate the standardized chemical information based on the Tanimoto similarity calculation method to obtain the calculation result, and combine the existing classification data with the calculation result to preliminarily predict the classification of the target pollutant.

[0145] The similarity coefficient of the Tanimoto similarity calculation method ranges from 0 to 1, and the closer the similarity coefficient is to 1, the higher the similarity is. When the similarity coefficient is greater than 0.8, it can be considered as highly similar. When integrated with weighting, a linear weighting method can be used, and the weight can be determined by grid search.

[0146] Step 3, simulate the migration path of the target pollutant in the environment using a graph neural network model, and analyze the migration, transformation and degradation behavior of the target pollutant in different environmental media to generate analysis results.

[0147] The graph neural network model adopts a graph convolutional neural network model based on an attention mechanism. The model input is pollutant molecular structure data and environmental parameter data, and the output is the migration path and transformation and degradation of the pollutant in different environmental media. Environmental media can include atmosphere, water body, soil, etc.

[0148] Step 4, according to the analysis results, comprehensively consider the persistence, mobility and bioaccumulation of the target pollutant, and calculate the comprehensive mechanism characteristic score of the target pollutant by weighted summation for multi-mechanism comprehensive evaluation.

[0149] The persistence score ranges from 0 to 1, the mobility score ranges from 0 to 1, and the bioaccumulation score ranges from 0 to 1. The sum of the weighting coefficients is 1. The weighting coefficients can be set to 0.4, 0.3 and 0.3 respectively.

[0150] Step 5, introduce an environmental factor correction model to correct the intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity and light intensity. The environmental factor correction model adopts a nonlinear regression model based on random forest, and the input is environmental parameter data and the output is a correction coefficient. Environmental parameter data can include temperature range of 10-35℃, pH value range of 5-9, and dissolved oxygen concentration range of 2-10mg / L.

[0151] Step 6, integrate the molecular structure, migration path, mechanism characteristics and environmental correction results to calculate the final classification prediction value of the target pollutant. The integration is performed by weighted summation, and the weights are set to 0.3, 0.3, 0.2 and 0.2 respectively. The weights are determined by 5-fold cross-validation.

[0152] Step 7, according to the final classification prediction value, divide the risk level of the target pollutant, and output the classification result. The risk level is divided into three levels: high risk, medium risk and low risk, and the corresponding classification prediction value ranges are 0.7-1, 0.4-0.7 and 0-0.4 respectively.

[0153] The embodiment realizes: collecting and standardizing chemical information: collecting chemical information of the target pollutant, including molecular structure description, molecular fingerprint, and physicochemical properties, and standardizing the information. The attention mechanism is used to find the substructure that contributes greatly to the predicted attribute value, and the parameter migration is used to solve the problem of insufficient sample size. The interspecies extrapolation normalization model is established, the number of species in the derivation of the soil environmental baseline value is increased, so that the derived baseline value is more scientific and reliable.

[0154] In addition, the embodiment of the application further provides a computer readable storage medium, and the storage medium stores a new pollutant classification prediction program. When the new pollutant classification prediction program is executed by a processor, the steps of the method for predicting the classification of the new pollutant are realized.

[0155] Reference Figure 3 , Figure 3 The structure block diagram of the first embodiment of the new pollutant classification prediction system of the application is shown in the figure.

[0156] As Figure 3 shown, the new pollutant classification prediction system provided by the embodiment of the application comprises:

[0157] An information collection module 10 is configured to collect chemical information of a target pollutant, wherein the chemical information comprises: molecular structure description, molecular fingerprint, and physicochemical properties, and the chemical information is standardized.

[0158] A preliminary classification evaluation result module 20 is configured to calculate the standardized chemical information based on a Tanimoto similarity calculation method to obtain a calculation result, and combine the calculation result with existing classification data to preliminarily predict the classification of the target pollutant and generate a preliminary classification evaluation result.

[0159] An analysis result module 30 is configured to simulate the migration path of the target pollutant in the environment by using a graph neural network, and analyze the migration, transformation, and degradation behavior of the target pollutant in different environmental media to generate an analysis result.

[0160] A comprehensive mechanism characteristic score module 40 is configured to comprehensively consider the persistence, migration, and bioaccumulation of the target pollutant according to the analysis result, and calculate the comprehensive mechanism characteristic score of the target pollutant by weighted summation for multi-mechanism comprehensive evaluation.

[0161] An intrinsic characteristic score module 50 is configured to introduce an environmental factor correction model, and correct the intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity, and light intensity.

[0162] An ultimate classification prediction value module 60 is configured to calculate an ultimate classification prediction value of the target pollutant based on the preliminary classification evaluation result, the analysis result, the intrinsic characteristic score and the environmental correction result.

[0163] An output module 70 is configured to divide the risk level of the target pollutant according to the ultimate classification prediction value and output a classification result.

[0164] It should be understood that the above is only an example, and does not limit the technical solutions of the present application. In specific applications, those skilled in the art can set up according to the needs, and the present application does not limit this.

[0165] The embodiment collects the chemical information (molecular structure, molecular fingerprint, physicochemical property) of the target pollutant and performs standardization processing; calculates a preliminary classification based on Tanimoto similarity; simulates the migration path and transformation behavior of the pollutant in the environment by using a graph neural network; comprehensively considers the persistence, mobility and bioaccumulation for multi-mechanism comprehensive evaluation; introduces an environmental factor correction model to correct the score according to parameters such as temperature, pH value and dissolved oxygen concentration; integrates the molecular structure, migration path, mechanism characteristics and environmental correction result to calculate the ultimate classification prediction value; and divides the risk level (high, medium and low risk) according to the prediction value. Through machine learning technology, the pollutant risk is comprehensively evaluated, the evaluation accuracy and reliability are improved, and a scientific basis is provided for pollution prevention and control.

[0166] It should be noted that the above-described workflow is only illustrative and does not limit the scope of protection of the present application. In actual applications, those skilled in the art can select part or all of them to achieve the purpose of the embodiment scheme according to actual needs, which is not limited here.

[0167] In addition, technical details not described in detail in the embodiment can be referred to the method for predicting classification of new pollutants provided by any embodiment of the present application, which will not be described here.

[0168] In addition, it should be noted that in this paper, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or system. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or system including the element.

[0169] The above sequence numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0170] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, also can be through hardware, but in many cases the former is the better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of the contribution to the prior art can be embodied in the form of software products, the computer software product is stored in a storage medium (such as read-only memory (ReadOnly Memory, ROM) / RAM, disk, optical disc), including a number of instructions to make a terminal device (may be a mobile phone, computer, server, or network equipment, etc.) executes the method of each embodiment of the present application. The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, any equivalent structure or equivalent process transformation made by using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method of predicting new pollutant classification, characterized by, The method comprises the following steps: S1: Collecting chemical information of the target pollutant, the chemical information comprising molecular structure description, molecular fingerprint and physicochemical properties, and performing standardization processing on the chemical information; S2: Calculating the chemical information subjected to the standardization processing based on a Tanimoto similarity calculation method to obtain a calculation result, and performing weighted integration on the calculation result in combination with existing classification data to preliminarily predict the classification of the target pollutant and generate a preliminary classification evaluation result; S3: Simulating the migration path of the target pollutant in the environment by using a graph neural network and analyzing the migration, transformation and degradation behavior of the target pollutant in different environmental media to generate an analysis result; S4: Considering the persistence, migration and bioaccumulation of the target pollutant based on the analysis result, performing multi-mechanism comprehensive evaluation by weighted summation to calculate the comprehensive mechanism characteristic score corresponding to the target pollutant; S5: Introducing an environmental factor correction model, and correcting the intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity and light intensity; S6: Calculating the final classification prediction value of the target pollutant based on the preliminary classification evaluation result, the analysis result, the intrinsic characteristic score and the environmental correction result; S7: Dividing the risk level of the target pollutant according to the final classification prediction value and outputting the classification result; In the step S5, the environmental correction model is calculated by the following formula: wherein, is an environmental correction factor, intrinsic properties of the pollutant are scored as ; The environmental parameters are denoted by , is the score of the pollutant intrinsic properties; In the step S5, the final classification prediction value is calculated by the following formula: wherein represents the preliminary classification evaluation results based on the molecular structure, : represents the analysis results considering the migration path and conversion behavior, represents the intrinsic property score evaluated based on a plurality of classification mechanisms, represents the classification property after environmental correction; Weight parameters All satisfy: 。 2. The method of claim 1, wherein, The step of collecting the chemical information of the target pollutant, the chemical information comprising molecular structure description, molecular fingerprint and physicochemical properties, and performing standardization processing on the chemical information comprises the following steps: Chemical information of the target pollutant is collected, the molecular structure description in the chemical information is represented as a set of atoms consisting of a series of atoms: each atom with its unique chemical properties ; Converting the molecule into a fixed-dimensional vector by using MACCS or Morgan fingerprint, and setting the molecular fingerprint of F as a vector; wherein each component denotes the presence or absence of a certain structure or functional group; standardized to represent physico-chemical properties : wherein and are the mean and standard deviation of the feature, respectively.

3. The method of claim 1, wherein, In the step S2, the Tanimoto similarity calculation formula is as follows: where the molecular fingerprint of the target pollutant is , the molecular fingerprint of the th known classified compound in the database is , the inner product is , and the vector norm is .

4. The method of claim 1, wherein, In the step S3, the migration path analysis is realized by using a graph neural network, and specifically comprises the following steps: Dividing the environment into multiple nodes, each node representing a specific environmental medium, and the characteristics of each node including the migration rate, adsorption capacity and degradation rate of the pollutant; Modeling the migration path of the pollutant by using a graph neural network, the input of the network including the physicochemical properties of the pollutant and environmental factors, so as to calculate the migration probability between different environmental nodes, and output the migration path and expected residence time of the pollutant in the environment.

5. The method of claim 1, wherein, In the step S4, the comprehensive mechanism characteristic score is calculated by the following formula: wherein, , , respectively represent the weight of the persistence, mobility, and bioaccumulation mechanisms, represents a pollutant, , and respectively represent the score of the pollutant in terms of persistence, mobility, and bioaccumulation, represents the comprehensive mechanism characteristic score obtained based on the comprehensive evaluation of the three classification mechanisms of persistence, mobility, and bioaccumulation.

6. A new pollutant classification prediction system characterized by, The method is executed as claimed in claim 1, comprising: An information collection module for collecting chemical information of the target pollutant, the chemical information comprising molecular structure description, molecular fingerprint and physicochemical properties, and performing standardization processing on the chemical information; A preliminary classification evaluation result module for calculating the chemical information subjected to the standardization processing based on a Tanimoto similarity calculation method to obtain a calculation result, and performing weighted integration on the calculation result in combination with existing classification data to preliminarily predict the classification of the target pollutant and generate a preliminary classification evaluation result; The analysis result module is configured to simulate a migration path of the target pollutant in an environment by using a graph neural network and analyze migration, transformation and degradation behaviors of the target pollutant in different environmental media to generate an analysis result. The comprehensive mechanism characteristic score module is configured to comprehensively consider persistence, migration and bioaccumulation of the target pollutant according to the analysis result, and calculate a comprehensive mechanism characteristic score of the target pollutant by weighted summation for multi-mechanism comprehensive evaluation. The intrinsic characteristic score module is configured to introduce an environmental factor correction model, and correct an intrinsic characteristic score of the target pollutant according to temperature, pH value, dissolved oxygen concentration, humidity and light intensity. The final classification prediction value module is configured to calculate a final classification prediction value of the target pollutant based on the preliminary classification evaluation result, the analysis result, the intrinsic characteristic score and the environmental correction result. The output module is configured to divide a risk level of the target pollutant according to the final classification prediction value, and output a classification result.

7. A computer device, characterized by The device comprises a memory and a processor, and the processor executes the method according to any one of claims 1 to 5 when running computer instructions stored in the memory.

8. A computer-readable storage medium, characterized in that, The instructions, when running on a computer, cause the computer to execute the method according to any one of claims 1 to 5.

Images