Method and system for evaluating acute toxicity of environmental pollutants by using zebra fish and storage medium

Abstract

The invention provides a method and a system for evaluating acute toxicity of environmental pollutants by using zebra fish and a storage medium, and relates to the technical field of pollutant toxicity evaluation, a behavior characteristic and molecular characteristic model is constructed, a toxicity evaluation model is generated in combination with multi-modal data, and comprehensive toxicity evaluation from behavior to molecular mechanism is realized. According to the method, multi-dimensional features of pollutants are extracted by using a deep learning technology, and rapid prediction of the influence of pollutant concentration on toxicity is realized through a concentration-toxicity mapping model, so that the precision and comprehensiveness of toxicity evaluation are improved, the use of zebra fish in an experiment is effectively reduced, and the experiment cost is reduced. Besides, in subsequent detection, the toxicity value can be rapidly predicted only by collecting the pollutant concentration, the response speed of toxicity evaluation is greatly improved, the method meets the requirements of high-flux pollutant screening and environmental monitoring, and scientific and efficient technical support is provided for toxicity research and management of water pollution.

Description

Technical Field

[0001] The present invention relates to the technical field of pollutant toxicity evaluation, and specifically to a method, system and storage medium for evaluating the acute toxicity of environmental pollutants using zebrafish. Background Technique

[0002] Currently, the toxic effects of pollutants in water bodies on the ecosystem have become a key research area. Due to its biological characteristics, including rapid development, high similarity of the genome to humans, and low experimental cost, zebrafish are widely used in toxicity evaluation experiments. However, traditional toxicity evaluation methods have significant deficiencies: on the one hand, they often evaluate toxicity only through a single dimension (such as behavioral or molecular data), making it difficult to comprehensively reveal the multi-level effects of pollutants on organisms; on the other hand, they rely heavily on actual experiments, which have a long experimental period and consume a large number of animals, not only limiting the rapid evaluation and large-scale screening of pollutant toxicity, but also increasing the experimental cost. In addition, due to the individual differences and complexity of experimental data, the accuracy and reliability of the results are often interfered with.

[0003] In the prior art, the publication number CN117973219A discloses a method and system for evaluating the acute toxicity of environmental pollutants using zebrafish. Through digital virtual technology, a zebrafish model is digitally virtually constructed on a digital platform to obtain a digital zebrafish model; through a toxicity evaluation model embedded in the digital platform, based on the virtual behavior of the digital zebrafish model and the physical behavior of the zebrafish model, the acute toxicity of environmental pollutants in the water environment is evaluated to obtain an acute toxicity evaluation result of the environmental pollutants. Although it can solve the technical problems in the prior art that the toxicity detection of zebrafish is difficult and the intuitiveness is poor, affecting the detection accuracy of pollutant toxicity, the toxicity evaluation is mainly based on the behavioral characteristics of zebrafish and cannot reveal the toxicity mechanism of pollutants from a microscopic direction. Moreover, its virtual zebrafish model is constructed based on a non-toxic virtual water body. When it is embedded in a toxic virtual water body, the behavioral characteristics of the virtual zebrafish may not be the same as those of the actual zebrafish, which also reduces the overall prediction ability of the model.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, system and storage medium for evaluating the acute toxicity of environmental pollutants using zebrafish, so as to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] The specific steps of the method for evaluating the acute toxicity of environmental pollutants using zebrafish include:

[0008] S1: Design a zebrafish behavior experiment, collect the behavior data of zebrafish during the experiment and the body fluid data after the experiment, and process the behavior data and body fluid data respectively.

[0009] S2: Based on the processed behavior data and body fluid data, construct a behavior feature model and a molecular feature model respectively, and couple the behavior feature model and the molecular feature model to generate a toxicity evaluation model.

[0010] S3: Based on the processed behavior data and body fluid data, construct a behavior prediction model and a molecular prediction model respectively, and use the behavior prediction model and the molecular prediction model as the input ends of the toxicity evaluation model to generate the final predicted toxicity value.

[0011] Preferably, the design logic of the zebrafish behavior experiment is:

[0012] Divide healthy zebrafish into an embryo group, a juvenile fish group, and an adult fish group evenly. Place all zebrafish in pollutant solutions with different concentrations for the same time. The number of zebrafish in the embryo group, juvenile fish group, and adult fish group in each concentration of pollutant solution is equal, and set the concentration of one group of pollutant solution to 0 as a control.

[0013] Use a behavior tracking system to collect data on zebrafish to obtain the behavior data of zebrafish at different stages in pollutant solutions with different concentrations. The behavior data includes swimming speed, turning angle, activity range, acceleration, and circadian rhythm changes.

[0014] After the experiment, extract body fluid samples from each group of zebrafish to obtain the body fluid data of zebrafish at different stages after the experiment. The body fluid data includes the concentration of oxidative stress substances, the concentration of inflammatory factors, and the concentration of neurotransmitters.

[0015] Preferably, the logic for processing the behavior data and body fluid data is:

[0016] Use the behavior data and body fluid data respectively to construct a behavior data matrix and a body fluid data matrix, which are expressed as:

[0017]

[0018] In the formula, τ represents the pollutant concentration, τ min 、τ max represent the minimum and maximum values of the pollutant concentration respectively, represent the behavior data matrix and the body fluid data matrix corresponding to the qth zebrafish in the pth form at the concentration of τ respectively, respectively represent the behavioral data matrix and body fluid data matrix corresponding to the q-th zebrafish in the p-th form at all concentrations. p and q respectively represent the indices of zebrafish forms and individuals, p ∈ [1, 3], q ∈ [1, M], and M represents the total number of zebrafish individuals in this form; v i , θ i , ρ i , a i , d i respectively represent the swimming speed, turning angle, activity range, acceleration, and circadian rhythm change at the i-th collection; Co i , Cy i , Cs i respectively represent the concentration of oxidative stress substances, the concentration of inflammatory factors, and the concentration of neurotransmitters at the i-th collection. N represents the total number of collections;

[0019] Perform mean processing on the behavioral data matrix and body fluid data matrix corresponding to the individuals of zebrafish in each form. The calculation method is:

[0020]

[0021]

[0022] In the formula respectively represent the behavioral data matrix and body fluid data matrix after mean processing;

[0023] Then perform standardization processing on each type of matrix element in the behavioral data matrix and body fluid data matrix after mean processing;

[0024] Finally, construct an observation window based on experimental requirements and use the sliding window method to smooth each type of matrix element in the standardized behavioral data matrix and body fluid data matrix.

[0025] Preferably, the behavioral feature model is constructed using a convolutional neural network, and the behavioral data matrix after data processing is input into the convolutional neural network to extract the corresponding behavioral features. The calculation method is:

[0026]

[0027] In the formula represents the behavioral feature vector of zebrafish in the p-th form, and CNN represents the convolutional neural network function;

[0028] The molecular feature model is constructed using a recurrent neural network, and the body fluid data matrix after data processing is input into the recurrent neural network to extract the corresponding molecular features. The calculation method is:

[0029]

[0030] In the formula represents the molecular feature vector of zebrafish in the p-th form, and RNN represents the recurrent neural network function.

[0031] Preferably, the logic for constructing the toxicity evaluation model is as follows:

[0032] Concatenate the behavioral features and molecular features to generate a joint feature vector, and the calculation method is:

[0033]

[0034] In the formula, H p represents the joint feature vector, represents the concatenation operation;

[0035] Input the joint feature vector into a multi-layer perceptron to generate the predicted acute toxicity value, and the calculation method is:

[0036] T y = MLP(H p )

[0037] In the formula, T y represents the predicted acute toxicity value, and MLP represents the multi-layer perceptron function.

[0038] Preferably, use the mean squared error as the loss function of the multi-layer perceptron function, and the loss function is expressed as:

[0039]

[0040] In the formula, F loss represents the loss function, T y (p,q) represents the predicted acute toxicity value according to the data of the q-th zebrafish in the p-th form, and T true represents the preset true toxicity value.

[0041] Preferably, the construction logic of the behavior prediction model and the molecular prediction model is as follows:

[0042] According to the behavioral feature vector and the molecular feature vector, respectively fit the mapping relationships of both with respect to the pollutant concentration, and the mapping relationships are expressed as:

[0043]

[0044] In the formula respectively represent the behavioral feature vector and the molecular feature vector at the pollutant concentration of τ, and f b , f t respectively represent the mapping equations between the pollutant concentration and the behavioral feature vector and the molecular feature vector.

[0045] Preferably, the logic for generating the final predicted toxicity value is as follows:

[0046] After collecting the pollutant concentration, substitute it into the mapping equations between the pollutant concentration and the behavioral feature vector and the molecular feature vector respectively to obtain the predicted behavioral feature vector and molecular feature vector.

[0047] Concatenate the predicted behavioral feature vector and molecular feature vector to generate a predicted combined feature vector, and then input the predicted second-generation combined feature vector into a multi-layer perceptron to obtain the predicted acute toxicity value corresponding to the pollutant concentration.

[0048] A system for evaluating the acute toxicity of environmental pollutants using zebrafish, the system for evaluating the acute toxicity of environmental pollutants is used to execute the above method for evaluating the acute toxicity of environmental pollutants, including:

[0049] A storage medium for storing a computer program;

[0050] A processor for executing the computer program to implement the monitoring and warning method according to any one of claims 1-8.

[0051] A storage medium, the storage medium is used to store a computer program, and the computer program realizes the above method for evaluating the acute toxicity of environmental pollutants when executed by a processor.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] By constructing a behavioral feature and molecular feature model and combining multi-modal data to generate a toxicity evaluation model, the present invention realizes a comprehensive toxicity assessment from behavior to molecular mechanism. Using deep learning technology to extract multi-dimensional features of pollutants and realizing the rapid prediction of the impact of pollutant concentration on toxicity through a concentration-toxicity mapping model not only improves the accuracy and comprehensiveness of toxicity evaluation, but also effectively reduces the use of zebrafish in experiments and lowers the experimental cost. In addition, in subsequent detections, only the pollutant concentration needs to be collected to quickly predict the toxicity value, greatly improving the response speed of toxicity evaluation and meeting the needs of high-throughput pollutant screening and environmental monitoring, providing scientific and efficient technical support for the toxicity research and management of water pollution. Description of the Drawings

[0054] Figure 1 It is a schematic diagram of the overall method flow of the present invention. Detailed Embodiments

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further details the present invention in conjunction with specific embodiments.

[0056] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0057] Example:

[0058] Please refer to Figure 1 , the present invention provides a technical solution:

[0059] A method for evaluating the acute toxicity of environmental pollutants using zebrafish, the specific steps including:

[0060] S1: Design a zebrafish behavior experiment, collect the behavior data of zebrafish during the experiment and the body fluid data after the experiment, and perform data processing on the behavior data and the body fluid data respectively.

[0061] The design logic of the zebrafish behavior experiment is:

[0062] Evenly divide healthy zebrafish into an embryo group, a larval group, and an adult group. Place all zebrafish in pollutant solutions of different concentrations for the same time. The number of zebrafish in the embryo group, larval group, and adult group in each concentration of pollutant solution is equal, and set the concentration of one group of pollutant solution to 0 as a control;

[0063] Use a behavior tracking system to collect data on zebrafish, and obtain the behavior data of zebrafish in pollutant solutions of different concentrations at different stages. The behavior data includes swimming speed, turning angle, activity range, acceleration, and circadian rhythm changes;

[0064] After the experiment, extract body fluid samples from each group of zebrafish to obtain the body fluid data of zebrafish after the experiment. The body fluid data includes the concentrations of oxidative stress substances, inflammatory factors, and neurotransmitters.

[0065] In this embodiment, it is assumed that a pollutant solution with a pollutant concentration range between 0 mg / L and 10 mg / L is used to set up 10 experimental groups and a control group at intervals of 1 mg / L. Each group includes 10 zebrafish embryos, 10 zebrafish larvae, and 10 zebrafish adults. The experimental time is 72 h, and behavioral data is collected once per minute to intuitively reflect the impact of pollutants on their nervous systems and motor abilities. After the experiment, liquid chromatography-tandem mass spectrometry (LC-MS / MS) is used to measure the concentrations of oxidative stress substances, inflammatory factors, and neurotransmitters. Specifically, glutathione (GSH) and hydrogen peroxide (H2O2) can be selected as markers for measuring oxidative stress substances, interleukin (IL-6) and tumor necrosis factor (TNFα) can be selected as markers for measuring inflammatory factors, and dopamine (DA) and acetylcholine (ACh) can be selected as markers for measuring neurotransmitters.

[0066] In the experimental design, including zebrafish at different developmental stages (embryos, larvae, adults) helps to comprehensively evaluate the differential effects of pollutant toxicity on different growth stages. The concentration gradient can quantitatively reveal the dose-effect relationship between pollutant concentration and toxicity effects, and the setting of the control group provides a baseline reference, which helps to clearly distinguish the degree of pollutant impact. Among them, behavioral data can reflect the comprehensive impact of pollutants on the nervous system and motor abilities of zebrafish, and body fluid data can reveal the impact of pollutants on the internal environment of organisms from the molecular mechanism level. By combining behavioral data and molecular data, a comprehensive toxicity assessment from macro to micro can be achieved, thereby constructing a more accurate toxicity model and improving the reliability of toxicity evaluation.

[0067] The logic for data processing of behavioral data and body fluid data is as follows:

[0068] The behavioral data matrix and body fluid data matrix are constructed using behavioral data and body fluid data respectively, and they are expressed as:

[0069]

[0070] In the formula, τ represents the pollutant concentration, τ min and τ max represent the minimum and maximum values of the pollutant concentration respectively, represent the behavioral data matrix and body fluid data matrix corresponding to the qth zebrafish in the pth form at the concentration of τ respectively, represent the behavioral data matrix and body fluid data matrix corresponding to the qth zebrafish in the pth form under all concentrations respectively. p and q represent the indices of zebrafish form and individual, p ∈ [1, 3], q ∈ [1, M], and M represents the total number of zebrafish individuals in this form; v i and θi , ρ i , a i , d i respectively represent the swimming speed, turning angle, activity range, acceleration, and circadian rhythm change at the i-th collection; Co i , Cy i , Cs i respectively represent the concentration of oxidative stress substances, concentration of inflammatory factors, and concentration of neurotransmitters at the i-th collection. N represents the total number of collections. The construction of the data matrix can clarify the organization form of the data and provide a standardized structure for the input of subsequent machine learning models;

[0071] Perform mean processing on the behavioral data matrix and body fluid data matrix corresponding to each individual zebrafish in each morphology. The calculation method is as follows:

[0072]

[0073] In the formula respectively represent the behavioral data matrix and body fluid data matrix after mean processing. Perform mean processing on the behavioral data and body fluid data of zebrafish in each morphology (embryo, larva, adult), which can effectively reduce the interference of individual differences on the experimental results and make the experimental results more representative.

[0074] Then perform standardization processing on each type of matrix element in the behavioral data matrix and body fluid data matrix after mean processing.

[0075] When performing standardization processing, various different normalization methods can be selected according to actual needs to eliminate the influence of data dimensions. In this embodiment, taking the swimming speed as an example, the specific standardization processing can be calculated using the following equation:

[0076]

[0077] v in the formula i ′ represents the swimming speed after standardization processing, σ respectively represent the mean and standard deviation of the swimming speed collected for this zebrafish individual in the experiment. The calculation methods of the two are as follows:

[0078]

[0079] The normalization methods for these data, including turning angle, activity range, acceleration, circadian rhythm changes, concentrations of oxidative stress substances, inflammatory factors, and neurotransmitters, are the same as those for swimming speed. Through normalization, the influence of data dimensions is eliminated, enabling different characteristics of behavioral data and body fluid data to be compared and coupled on the same scale. Taking swimming speed as an example, data normalization not only retains the relative change trends of each data characteristic but also eliminates the potential impact of absolute numerical differences on the modeling results, improves the training efficiency of subsequent models, and reduces the problem of gradient instability caused by numerical range differences. At the same time, it allows data with different feature dimensions to be effectively fused, facilitating the construction and interpretation of toxicity models.

[0080] Finally, an observation window is constructed based on experimental requirements, and the sliding window method is used to smooth each type of matrix element in the normalized behavioral data matrix and body fluid data matrix. The specific calculation method is expressed as:

[0081]

[0082] where v i1 ′′ represents the data corresponding to the i1 - th acquisition after smoothing, and w represents the observation window. Smoothing extracts the main trend of the data, can better reflect the overall impact of pollutant concentration on toxicity characteristics, improves the robustness of toxicity model prediction, and reduces the error caused by noise.

[0083] In this step, through the acquisition of both behavioral data and molecular data, the toxicity assessment model can simultaneously consider the macroscopic impact (behavior) and microscopic mechanism (molecular level) of pollutants, improving the accuracy of toxicity prediction. Moreover, by constructing a logical chain of matrix - mean normalization - standardization - smoothing, the data becomes more standardized, providing a standardized format for the input of subsequent machine learning models.

[0084] S2: Based on the processed behavioral data and body fluid data, construct a behavioral feature model and a molecular feature model respectively, and couple the behavioral feature model and the molecular feature model to generate a toxicity assessment model.

[0085] The behavioral feature model is constructed using a convolutional neural network, and the processed behavioral data matrix is input into the convolutional neural network to extract the corresponding behavioral features. The calculation method is:

[0086]

[0087] In the formula represents the behavioral feature vector of zebrafish in the p - th form, and CNN represents the convolutional neural network function.

[0088] Convolutional neural network (CNN) is a model that is good at processing spatiotemporal data (such as behavioral data matrix). It extracts local features in the data through convolution and pooling operations, and its output behavioral feature vector can reflect the impact of pollutant concentration on zebrafish behavior. When used in this embodiment, a three-layer convolutional neural network can be used for processing. The convolution kernel size of the first convolution layer is set to 3X3, which is used to extract local features. Features related to pollutant concentration in behavioral data can be extracted through the local receptive field, such as behavioral change patterns within a certain time window; the pooling window of the second pooling layer is set to 2X2, which is used to reduce the data dimension, can compress the data dimension, reduce the amount of calculation, improve the model efficiency, and retain key features; the third layer is a fully connected layer, which is used to output the behavioral feature vector, because the convolution kernel can learn the potential laws and patterns in the behavioral data, and generate a feature vector that can characterize the impact of pollutants on zebrafish behavior. Since convolutional neural networks are mature existing technologies, they are not described here.

[0089] The behavioral feature vector can reflect the effect of pollutant concentration on zebrafish behavior and provide a direct representation of the nervous system and motor ability for toxicity evaluation. Moreover, the CNN structure is mature and easy to optimize, and is compatible with large-scale data, ensuring the efficiency and accuracy of feature extraction.

[0090] The molecular feature model is constructed using a recurrent neural network, and the body fluid data matrix after data processing is input into the recurrent neural network to extract the corresponding molecular features. The calculation method is:

[0091]

[0092] In the formula represents the molecular feature vector of zebrafish in the pth morphology, and RNN represents the recurrent neural network function.

[0093] Recurrent neural networks (RNNs) focus on processing time series data (such as changes in molecular concentrations in body fluid data matrices), and can capture the dynamic pattern of molecular toxicity changes over time. The molecular feature vectors they output can reflect the toxic effects of pollutants on molecular mechanisms. In this embodiment, a two-layer recurrent neural network can be used for processing. The size of the hidden layer in the first recurrent layer is set to 64 to capture time dependency, and to memorize the input information at previous moments, thereby capturing the effect of pollutant concentrations on molecular features over time; the second fully connected layer is used to output molecular feature feature vectors, because through the recursive processing of time steps, RNNs can gradually learn the effect of pollutant concentrations on the dynamic changes of molecular environments (such as oxidative stress and inflammatory responses). Since recurrent neural networks are also mature existing technologies, they will not be elaborated on here.

[0094] Molecular feature vectors can reflect the dynamic changes of the toxic effects of pollutants at the molecular level, provide in-depth explanations from the biochemical level, and introduce time series analysis, which can reveal the trend of toxic effects over time, making toxicity prediction more accurate.

[0095] The logic for constructing the toxicity evaluation model is as follows:

[0096] Concatenate the behavioral features and molecular features to generate a combined feature vector. The calculation method is as follows:

[0097]

[0098] In the formula, H p represents the combined feature vector, represents the concatenation operation. Concatenating the behavioral feature vector and the molecular feature vector to generate a combined feature vector comprehensively considers the dual effects of pollutant toxicity on behavior and molecular mechanisms, enabling the toxicity evaluation model to have the ability to input multi-modal features and more comprehensively reflect the toxic effects of pollutants.

[0099] Input the combined feature vector into a multi-layer perceptron to generate the predicted acute toxicity value. The calculation method is as follows:

[0100] T y = MLP(H p )

[0101] In the formula, T y represents the predicted acute toxicity value, and MLP represents the multi-layer perceptron function. MLP is a classic non-linear regression model. By introducing non-linear activation functions (such as ReLU, Sigmoid), MLP can capture the relationship between complex features and target values, thus mapping the combined feature vector to the predicted acute toxicity value.

[0102] Use the mean square error as the loss function of the multi-layer perceptron function. The loss function is expressed as:

[0103]

[0104] In the formula, F loss represents the loss function, T y (p,q) represents the predicted acute toxicity value based on the data of the qth zebrafish in the pth form, T trueRepresents the preset true toxicity value. The true toxicity value here can be obtained in various ways. Specifically, if the pollutant is a known existing substance, then its true toxicity value can directly refer to the toxicity value in the international toxicity standard library as the true toxicity value. For example, USEPA (U.S. Environmental Protection Agency): Environmental Toxicity Database, OECD (Organization for Economic Cooperation and Development): Chemical Toxicity Assessment Data, and China National Environmental Protection Standards (HJ series standards). If it is an unknown pollutant, then its true toxicity value can be set as the reciprocal of the concentration when 50% of the zebrafish die, that is, the reciprocal of the median lethal concentration (LC50), or it can be obtained by weighting based on the change amounts of the behavioral characteristic data and molecular characteristic data of zebrafish in the experimental group and the control group. All in all, its specific numerical value or calculation method can be obtained according to actual needs.

[0105] Using the mean squared error (MSE) as the loss function can quantify the error between the predicted value and the true value and guide the model to be gradually optimized. The loss function takes into account the toxicity evaluation results of different zebrafish morphologies (embryos, juveniles, adults), ensuring the applicability of the model to different developmental stages.

[0106] In this step, by introducing a deep learning model, it is possible to extract the characteristics of pollutant toxicity on behavior and molecular mechanisms from multiple dimensions, and generate a toxicity evaluation model through multi-modal data coupling to comprehensively evaluate pollutant toxicity from the macroscopic behavior level to the microscopic molecular mechanism level. This not only improves the accuracy and comprehensiveness of toxicity prediction but also provides the possibility for personalized and multi-scenario toxicity evaluation.

[0107] S3: Based on the processed behavioral data and body fluid data, construct a behavior prediction model and a molecular prediction model respectively, and use the behavior prediction model and the molecular prediction model as the input ends of the toxicity evaluation model to generate the final predicted toxicity value.

[0108] The construction logic of the behavior prediction model and the molecular prediction model is as follows:

[0109] According to the behavioral feature vector and the molecular feature vector, respectively fit the mapping relationship between the two and the pollutant concentration. The mapping relationship is expressed as:

[0110]

[0111] In the formula respectively represent the behavioral feature vector and the molecular feature vector at the pollutant concentration of τ, and f b 、f t respectively represent the mapping equations between the pollutant concentration and the behavioral feature vector and the molecular feature vector.

[0112] Here, flexible fitting methods (such as linear regression, polynomial regression, or deep neural network fitting) are used to capture the complex relationships between pollutant concentrations and behavioral and molecular characteristics. These two aspects represent the independent effects of pollutants on behavior and molecular mechanisms respectively. By combining the two, a comprehensive toxicity assessment from macroscopic behavior to microscopic molecular mechanisms can be achieved.

[0113] The logic for generating the final predicted toxicity value is as follows:

[0114] After collecting the pollutant concentration, substitute it into the mapping equations between the pollutant concentration and the behavioral and molecular feature vectors respectively to obtain the predicted behavioral and molecular feature vectors.

[0115] Concatenate the predicted behavioral and molecular feature vectors to generate a predicted combined feature vector, and then input the predicted second-generation combined feature vector into a multi-layer perceptron to obtain the predicted acute toxicity value corresponding to the pollutant concentration.

[0116] In this step, by constructing the behavioral prediction model and the molecular prediction model, the pollutant concentration, behavioral characteristics, and molecular characteristics are successfully coupled. This can not only explain the toxicity mechanism of pollutants from different levels, but also, during subsequent detection, only the pollutant concentration needs to be detected to predict its impact on the behavior and molecular levels of zebrafish, without the need to conduct experiments using actual zebrafish. This can not only more quickly understand the toxicity mechanism (such as the impact on behavior may not be obvious at low concentrations, but significant neurotoxic effects may be shown at high concentrations) and toxicity intensity of pollutants at this concentration, but also reduce the consumption of zebrafish, achieving the effects of improving response and reducing costs.

[0117] In summary, the present invention realizes a comprehensive toxicity assessment from behavior to molecular mechanisms by constructing behavioral and molecular feature models and combining multi-modal data to generate a toxicity evaluation model. Using deep learning technology to extract multi-dimensional features of pollutants and realizing rapid prediction of the impact of pollutant concentration on toxicity through a concentration-toxicity mapping model not only improves the accuracy and comprehensiveness of toxicity evaluation, but also effectively reduces the use of zebrafish in experiments and lowers the experimental cost. In addition, in subsequent detections, only the pollutant concentration needs to be collected to quickly predict the toxicity value, greatly improving the response speed of toxicity evaluation and meeting the requirements of high-throughput pollutant screening and environmental monitoring, providing scientific and efficient technical support for the toxicity research and management of water pollution.

[0118] The present invention also provides a system for evaluating the acute toxicity of environmental pollutants using zebrafish, which is used to execute the method for evaluating the acute toxicity of environmental pollutants as described above, including:

[0119] A storage medium for storing a computer program;

[0120] A processor for executing a computer program to implement the above-mentioned monitoring and warning method.

[0121] The present invention also provides a storage medium for storing a computer program, and when the computer program is executed by a processor, the method for evaluating the acute toxicity of environmental pollutants is implemented.

[0122] The above formulas are all dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0123] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0124] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0125] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all of them should be covered by the protection scope of this application.

Claims

1. A method for evaluating the acute toxicity of environmental pollutants using zebrafish, characterized in that: The specific steps include: S1: Design a zebrafish behavior experiment, collect the zebrafish behavior data during the experiment and the body fluid data after the experiment, and process the behavior data and body fluid data respectively: S2: Based on the processed behavioral data and body fluid data, a behavioral characteristic model and a molecular characteristic model are constructed respectively, and the behavioral characteristic model and the molecular characteristic model are coupled to generate a toxicity evaluation model; S3: Based on the behavioral data and body fluid data after data processing, a behavioral prediction model and a molecular prediction model are constructed respectively, and the behavioral prediction model and the molecular prediction model are used as the input of the toxicity evaluation model to generate the final predicted toxicity value.

2. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 1, characterized in that: The design logic of the zebrafish behavior experiment is: Healthy zebrafish were divided into embryo group, larval group and adult group. All zebrafish were placed in pollutant solutions of different concentrations for the same time. The number of zebrafish in each pollutant solution was equal. The concentration of one pollutant solution was set to 0 as a control. Use a behavior tracking system to collect data on zebrafish, and obtain behavioral data of zebrafish at different stages in pollutant solutions with different concentrations. The behavioral data include swimming speed, turning angle, activity range, acceleration, and circadian rhythm changes; After the experiment, body fluid samples were extracted from each group of zebrafish to obtain body fluid data of zebrafish at different stages after the experiment, including the concentration of oxidative stressors, inflammatory factors, and neurotransmitters.

3. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 2, characterized in that: The logic for data processing of behavioral data and body fluid data is as follows: The behavioral data and body fluid data are used to construct the behavioral data matrix and the body fluid data matrix respectively, which are expressed as: Where τ represents the pollutant concentration, τ min , τ max represent the minimum and maximum values ​​of pollutant concentrations, respectively. They represent the behavior data matrix and body fluid data matrix corresponding to the qth zebrafish in the pth form when the concentration is τ, Respectively represent the behavioral data matrix and body fluid data matrix corresponding to the qth zebrafish in the pth morphology under all concentrations, p and q represent the index of the zebrafish morphology and individual, p∈[1,3], q∈[1,M], M represents the total number of zebrafish individuals in this morphology; v i ,θ i , i 、a i ,d i Respectively represent swimming speed, turning angle, activity range, acceleration, and circadian rhythm changes at the i-th collection; Co i , Cy i , Cs i They represent the concentrations of oxidative stressors, inflammatory factors, and neurotransmitters at the time of the i-th collection, respectively, and N represents the total number of collections; The behavioral data matrix and body fluid data matrix corresponding to each individual zebrafish in each morphology are averaged and calculated as follows: In the formula Respectively represent the behavioral data matrix and body fluid data matrix after mean processing; Then, each type of matrix element in the behavioral data matrix and body fluid data matrix after mean processing is standardized; Finally, an observation window is constructed based on experimental requirements, and the sliding window method is used to smooth each type of matrix element in the standardized behavioral data matrix and body fluid data matrix.

4. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 3, characterized in that: The behavior feature model is constructed using a convolutional neural network, and the behavior data matrix after data processing is input into the convolutional neural network to extract the corresponding behavior features. The calculation method is: In the formula represents the behavioral feature vector of zebrafish in the pth form, and CNN represents the convolutional neural network function; The molecular feature model is constructed using a recurrent neural network, and the body fluid data matrix after data processing is input into the recurrent neural network to extract the corresponding molecular features. The calculation method is: In the formula represents the molecular feature vector of zebrafish in the pth morphology, and RNN represents the recurrent neural network function.

5. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 4, characterized in that: The logic of constructing the toxicity evaluation model is: The behavioral features and molecular features are spliced ​​to generate a joint feature vector, which is calculated as follows: Where Hp represents the joint eigenvector, Represents a splicing operation; The joint feature vector is input into a multilayer perceptron to generate a predicted acute toxicity value, which is calculated as: T y =MLP(H p ) Where T y represents the predicted acute toxicity value, and MLP represents the multi-layer perceptron function.

6. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 5, characterized in that: Using mean square error as the loss function of the multilayer perceptron function, the loss function is expressed as: Where F loss represents the loss function, T y (p, q) represents the acute toxicity value predicted based on the data of the qth zebrafish in the pth morphology, T true Indicates the preset true toxicity value.

7. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 5, characterized in that: The construction logic of the behavior prediction model and the molecular prediction model is: According to the behavior feature vector and the molecular feature vector, the mapping relationship between the two on the pollutant concentration is fitted respectively, and the mapping relationship is expressed as: In the formula They represent the behavior characteristic vector and molecular characteristic vector when the pollutant concentration is τ, respectively, b 、f t Represent the mapping equations between pollutant concentration and behavioral characteristic vector and molecular characteristic vector respectively.

8. The method for evaluating the acute toxicity of environmental pollutants using zebrafish according to claim 7, characterized in that: The logic for generating the final predicted toxicity value is: After collecting the pollutant concentration, the pollutant concentration is substituted into the mapping equations between the behavior characteristic vector and the molecular characteristic vector to obtain the predicted behavior characteristic vector and molecular characteristic vector; The predicted behavioral feature vector and molecular feature vector are concatenated to generate a predicted joint feature vector, and then the predicted second-generation joint feature vector is input into a multi-layer perceptron to obtain the predicted acute toxicity value corresponding to the pollutant concentration.

9. A system for evaluating the acute toxicity of environmental pollutants using zebrafish, characterized in that: The system for evaluating the acute toxicity of environmental pollutants is used to perform the method for evaluating the acute toxicity of environmental pollutants according to any one of claims 1 to 8, comprising: Storage media for storing computer programs; A processor is used to execute the computer program to implement the monitoring and early warning method as described in any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium is used to store a computer program, and when the computer program is executed by a processor, the method for evaluating the acute toxicity of environmental pollutants according to any one of claims 1 to 8 is implemented.

Images