Water quality comprehensive toxicity high-flux determination method

Abstract

The invention discloses a water quality comprehensive toxicity high-throughput determination method, which comprises the following steps: performing exposure contamination on a tested organism by using a sewage sample to obtain phenotypic characteristic data of the tested organism; constructing a toxicity matrix; establishing a machine learning model, and determining the water quality comprehensive toxicity of the sewage sample in combination with the toxicity matrix. The water quality comprehensive toxicity determination method provided by the invention does not need sample concentration and enrichment, ensures the comprehensiveness and accuracy of the detection result, and can be applied to the comprehensive toxicity high-flux determination of the sewage in the whole process of a sewage treatment plant.

Description

Technical Field

[0001] The invention belongs to the field of water quality risk control, and specifically relates to a high-throughput determination method for comprehensive toxicity of water quality. Background Art

[0002] The wastewater generated in the process of human life and production has complex composition, great changes in water quality, and contains a large number of pollutants of different properties, which is one of the key factors causing water ecological security problems. With the gradual advancement of industrialization and urbanization, the pollutants generated by human life and production activities are increasing day by day. A large number of toxic and harmful pollutants with extremely low concentrations and high risks (such as antibiotics, microplastics and their derivatives, perfluorinated and polyfluoroalkyl substances, and new pollutants such as brominated flame retardants) enter the sewage, resulting in a significant increase in the potential risks of sewage, which has attracted widespread attention to the toxicity of sewage worldwide.

[0003] Since there are many kinds of pollutants in sewage, pollutants not only have direct toxic effects on organisms, but may also form complex toxic effects with other pollutants. It is difficult to accurately reflect the overall toxic effects of sewage by using the concentration level of a single / specific pollutant or the toxic effect value of a single / specific pollutant, which may produce large deviations. Commonly used methods for evaluating the overall toxic effects of sewage at home and abroad include the comprehensive toxicity method of wastewater, the toxicity identification assessment method, and the direct toxicity assessment method. Among them, the comprehensive toxicity assessment method of wastewater is to expose the standard model organism to a gradient dilution sample and determine the acute toxicity value of the sample to the test organism at a fixed exposure time. However, these methods require a lot of time to determine the acute toxicity value through multiple gradient dilution sample exposure experiments, and it is difficult to achieve rapid detection of the toxic effects of a large number of sewage samples in a short time. In addition, the test organisms used in toxicity tests have different sensitivities and tolerances to different pollutants / water quality, and the differences in sewage in different treatment process sections are extremely large. There is an urgent need for a high-throughput determination method for the comprehensive toxicity of sewage in the entire process of sewage treatment plants that is highly adaptable and sensitive. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a high-throughput determination method for the comprehensive toxicity of water quality, so as to improve the problems of low detection flux of existing methods for evaluating the overall toxic effects of sewage, differences in sensitivity and tolerance of the test organisms used in toxicity tests to different pollutants / water qualities, and great differences in sewage in different treatment process stages.

[0005] Technical solution: The high-throughput determination method for comprehensive water quality toxicity described in the present invention comprises the following steps: obtaining phenotypic characteristic data of the test organisms after exposing them to toxic substances using sewage samples; constructing a toxicity matrix; establishing a machine learning model, and determining the comprehensive toxicity of water quality of the sewage samples in combination with the toxicity matrix.

[0006] Preferably, the exposure time is 24 hours.

[0007] Preferably, the sewage samples are pretreated before being used to expose the test organisms to the poison.

[0008] Preferably, the sewage sample pretreatment process is to filter the sewage sample through a 0.22 μm water filter membrane.

[0009] Preferably, the test organisms are algae cells and fish gill cells.

[0010] Further preferably, the test organisms are cells of Capricornis solani and gill cells of rainbow trout.

[0011] Preferably, after the test organisms are exposed to the poison using sewage samples, the phenotypic characteristic data of the test organisms are obtained by using multiple fluorescent staining, high-content automatic imaging and cell morphological feature extraction.

[0012] Preferably, the multiple fluorescent stains used for algal cell staining consist of Hoechst 33342 dye, concanavalinA / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germagglutinin / Alexa Fluor 555 dyes; the multiple fluorescent stains used for fish gill cell staining consist of Hoechst 33342 dye, concanavalinA / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye, and MitoTracker Deep Red dye.

[0013] Preferably, the high-content automatic imaging uses a high-content cell imaging and analysis system to automatically collect subcellular structure images of algae cells and fish gill cells inoculated in 4 to 8 groups of parallel experiments on a well plate with high throughput.

[0014] Preferably, the image acquisition conditions of the high-content cell imaging and analysis system are as follows: 9 (3×3) imaging field points are set in each well of the well plate, 2×2 pixel merging is adopted, and each point automatically captures 5-color fluorescence channel images and 3 bright field channel images originating from different z-axis focal points, wherein a 63x water immersion objective is used to obtain the subcellular structure image of algae cells, and a 20x water immersion objective is used to obtain the subcellular structure image of fish gill cells; the excitation / emission wavelengths of the 5-color fluorescence channels for automatic imaging of algae cells are DNA 376-398nm / 417-477nm, ER 442-502nm / 503-538nm, RNA 491-571nm / 573-613nm, AGP 502-622nm / 622-662nm, Cy5588-668nm / 652-732nm; the excitation / emission wavelengths of the 5-color fluorescence channels for automatic imaging of fish gill cells are DNA 376~398nm / 417~477nm, ER 442~502nm / 503~538nm, RNA 491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Mito 588~668nm / 672~712nm.

[0015] Preferably, the step of extracting cell morphological features is: positioning the cells, cell nuclei, and cytoplasm of each image, and the image quality must meet the following conditions: the average image intensity is 10-240, the image focus score is >0.5, the image edge pre-degree standard deviation is <0.2, the cell area is 50-500, the number of cell debris holes is <5, the cell density is >50, and the cell nuclear staining clarity is >1.5; the gray level co-occurrence matrix algorithm of texture feature analysis is used to calculate the morphology, intensity, texture, brightness, average grayscale of each cell and the minimum distance, adjacency value, and aggregation degree between each cell, and obtain the cell morphological feature items of each cell and the arithmetic mean of the cell morphological feature values ​​corresponding to each cell morphological feature item.

[0016] Preferably, 5797 cell morphological characteristics are obtained for each cell.

[0017] Preferably, the toxicity matrix is ​​composed of the phenotypic characteristic data of algae cells and fish gill cells, which are filtered by feature items and normalized by feature values, and then clustered and arranged; the feature item filtering is to exclude feature items with collinear intersections and retain feature items with eigenvalues ​​not equal to 0; the eigenvalue standardization method is the Z-Score method and the maximum-minimum method; the cluster arrangement method is to classify and integrate the feature items according to the subcellular structures corresponding to the feature items, and the arrangement classification is an ultra-high dimensional feature item consisting of algae cell DNA, algae cell endoplasmic reticulum, algae cell nucleosomes and cytoplasmic RNA, algae cell actin and Golgi apparatus and plasma membrane, algae cell chloroplasts, algae cell bright field, fish gill cell DNA, fish gill cell endoplasmic reticulum, fish gill cell nucleosomes and cytoplasmic RNA, fish gill cell actin and Golgi apparatus and plasma membrane, fish gill cell mitochondria, and fish gill cell bright field.

[0018] Preferably, the machine learning model is based on the acute toxicity effect values ​​and phenotypic characteristic data of algal cells and the acute toxicity effect values ​​and phenotypic characteristic data of fish gill cells, and is established by one of a random forest model, an XGBoost algorithm model, a Lasso regression algorithm model, a content-based recommendation algorithm model, and a support vector machine model.

[0019] Preferably, the method for determining the comprehensive toxicity of water quality of sewage samples is: using the least partial squares discriminant analysis method to perform dimension reduction processing on the constructed toxicity matrix, obtain the characteristic variables of comprehensive toxicity of water quality, substitute them into the machine learning model to obtain the comprehensive toxicity of water quality of sewage samples; the comprehensive toxicity of water quality of sewage samples is the value of acute toxic effect caused by sewage samples, with the concentration that can cause 10% of the maximum effect (EC 10 )express.

[0020] Preferably, the comprehensive toxicity characteristic variables of water quality are 12 items.

[0021] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0022] (1) The present invention provides a method for high-throughput and rapid detection of the comprehensive toxicity of sewage water quality, which solves the problems of cumbersome detection steps, long time consumption and low detection throughput in existing methods.

[0023] (2) The present invention is based on high-content cell imaging technology and machine learning models to automatically capture the toxic effects of the comprehensive toxicity of sewage samples at the cellular and subcellular structure levels, and accurately quantify the acute toxic effect values ​​of sewage samples, thereby greatly simplifying the measurement steps and ensuring the comprehensiveness and accuracy of the test results.

[0024] (3) The present invention utilizes the different response characteristics of algal cells and fish gill cells to the toxic effects of sewage samples to construct a toxicity matrix, which reduces the selectivity of existing methods for specific sewage samples, enhances the universality of the high-throughput determination method of comprehensive water quality toxicity for various types of sewage samples, and improves the practical applicability of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The flowchart of the high-throughput determination method of comprehensive toxicity of water quality of the present invention is as follows;

[0026] Figure 2 The subcellular structure images of algae cells and fish gill cells in Example 1 of the present invention;

[0027] Figure 3 The toxicity matrix constructed according to the phenotypic characteristic data of algal cells and fish gill cells in Example 1 of the present invention;

[0028] Figure 4 The comprehensive toxicity of the water quality of the whole-process wastewater sample of Plant B in Example 2 of the present invention;

[0029] Figure 5 It is the comprehensive toxicity of the effluent from plants C, D and E in Example 3 of the present invention. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described below in conjunction with embodiments and drawings.

[0031] Example 1

[0032] The application object of this embodiment is the influent sample of a municipal sewage treatment plant in Jiangsu Province. The daily processing capacity of plant A is 80,000 cubic meters / day, and the influent contains 254.0 mg / L COD, 29.27 mg / L total nitrogen, and 2.07 mg / L total phosphorus. A high-throughput determination method for comprehensive toxicity of water quality, the specific establishment steps are:

[0033] Step 1: Filter the wastewater sample from Factory A through a 0.22μm water filter membrane.

[0034] Step 2: Using the cells of the ceratospermum elata and the gill cells of rainbow trout as the test organisms, the filtered sewage samples were used to expose the cells of the ceratospermum elata and the gill cells of rainbow trout for 24 hours; Hoechst 33342 dye, concanavalinA / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye were used to prepare multiple fluorescent stains, and the exposed algal cells were subjected to multiple fluorescent staining, using Hoechst 33342 dye, concanavalinA / AlexaFluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye, MitoTracker Deep Red dye was used to make multiple fluorescent stains, and multiple fluorescent staining was performed on the gill cells of fish exposed to the poison; a high-content cell imaging and analysis system was used to automatically collect subcellular structure images of algae cells and fish gill cells inoculated in 6 parallel experiments in the well plate: a 63x water-immersion objective was used to obtain subcellular structure images of algae cells, and a 20x water-immersion objective was used to obtain subcellular structure images of fish gill cells. The excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of algae cells were DNA 376~398nm / 417~477nm, ER 442~502nm / 503~538nm, RNA491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Cy5588~668nm / 652~732nm, and the excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of fish gill cells were DNA 376~398nm / 417~477nm, ER442~502nm / 503~538nm, RNA 491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Mito 588~668nm / 672~712nm, 9 (3×3) imaging field points are set in each well of the well plate, 2×2 pixel merging is used, and each point automatically captures 5-color fluorescence channel images and 3 bright field channel images from different z-axis focuses; the automatically captured images are used to extract cell morphological features, and the cells, cell nuclei, and cytoplasm of each image are located. The image quality must meet the following conditions: the average image intensity is 10~240, the image focus score value is >0.5, the image edge front standard deviation is <0.2, the cell area is 50~500, the number of cell debris holes is <5, the cell density is >50, and the clarity of cell nuclear staining is >1.5. The gray-level co-occurrence matrix algorithm of texture feature analysis was used to calculate the morphology, intensity, texture, brightness, average grayscale of each cell and the minimum distance, adjacency value, and aggregation degree between each cell, and the 5797 cell morphological feature items of each cell and the arithmetic mean of the cell morphological feature value corresponding to each cell morphological feature item were obtained.

[0035] Step 3: Exclude the collinear crossing feature items within the phenotypic characteristic data of algae cells and fish gill cells, and retain the feature items whose feature values ​​are not equal to 0; use the Z-Score method and the maximum-minimum method to standardize the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering; classify and integrate the feature items according to the subcellular structures corresponding to the feature items, and arrange and classify the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering and feature value standardization into ultra-high dimensional feature items consisting of algae cell DNA, algae cell endoplasmic reticulum, algae cell nucleosomes and cytoplasmic RNA, algae cell actin and Golgi apparatus and plasma membrane, algae cell chloroplasts, algae cell bright field, fish gill cell DNA, fish gill cell endoplasmic reticulum, fish gill cell nucleosomes and cytoplasmic RNA, fish gill cell actin and Golgi apparatus and plasma membrane, fish gill cell mitochondria, and fish gill cell bright field to construct a toxicity matrix.

[0036] Step 4: Based on the acute toxicity effect values ​​and phenotypic characteristic data of algal cells and the acute toxicity effect values ​​and phenotypic characteristic data of fish gill cells, a machine learning model was established through a random forest model; the constructed toxicity matrix was subjected to dimensionality reduction processing using the least partial squares discriminant analysis method to obtain 12 water quality comprehensive toxicity characteristic variables, which were substituted into the machine learning model to obtain the comprehensive water quality toxicity of the influent of Plant A, expressed as the concentration that can cause 10% maximum effect (EC10).

[0037] like Figure 2 As shown in Figure 1, the subcellular structure images of algae cells and fish gill cells of the influent sample of Plant A obtained by this method are shown in Figure 2. Figure 3 As shown in Table 1, the cell morphological characteristics were extracted to obtain the cell phenotypic characteristic data of algal cells and fish gill cells and construct a toxicity matrix. As shown in Table 1, the least partial squares discriminant analysis method was used to reduce the dimension of the toxicity matrix, and 12 water quality comprehensive toxicity characteristic variables were obtained. They were brought into the machine learning model to obtain the comprehensive toxicity of the water quality of Plant A's influent, which was 55.2%.

[0038] Table 1 12 water quality comprehensive toxicity characteristic variables obtained in Example 1

[0039] Comprehensive toxicity characteristic variables of water quality Numeric DNA_1 -20.173203 DNA_2 3.73463273 RNA_1 -15.385017 RNA_2 4.7597349 ER_1 -21.962306 ER_2 2.68094709 AGP_1 -18.088792 AGP_2 2.38816954 Chl -18.7958 Mito 4.47955748 BR_1 -17.150802 BR_2 7.18030231

[0040] Example 2

[0041] The difference from Example 1 is that the application object of this embodiment is the whole process sewage samples of a municipal sewage treatment plant in the southwest region, including sewage samples from the water inlet, aerated grit chamber, anoxic tank, aerobic tank, secondary sedimentation tank, sand filter, disinfection tank, and outlet. The daily processing capacity of Plant B is 450,000 cubic meters / day, and the inlet water contains 241.1 mg / L COD, 27.02 mg / L total nitrogen, and 2.94 mg / L total phosphorus; the outlet water contains 55.40 mg / L COD, 10.37 mg / L total nitrogen, and 0.38 mg / L total phosphorus. A high-throughput determination method for comprehensive toxicity of water quality, the specific establishment steps of which are:

[0042] Step 1: Filter the 8 wastewater samples from the entire process of Plant B through a 0.22μm water filter membrane.

[0043] Step 2: Using the cells of the ceratospermum elata and the gill cells of rainbow trout as the test organisms, the filtered sewage samples were used to expose the cells of the ceratospermum elata and the gill cells of rainbow trout for 24 hours; Hoechst 33342 dye, concanavalinA / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye were used to prepare multiple fluorescent stains, and the exposed algal cells were subjected to multiple fluorescent staining, using Hoechst 33342 dye, concanavalinA / AlexaFluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye, MitoTracker Deep Red dye was used to make multiple fluorescent stains, and multiple fluorescent staining was performed on the gill cells of fish exposed to the poison; a high-content cell imaging and analysis system was used to automatically collect subcellular structure images of algae cells and fish gill cells inoculated in 4 parallel experiments on the well plate with high throughput: a 63x water-immersion objective was used to obtain subcellular structure images of algae cells, and a 20x water-immersion objective was used to obtain subcellular structure images of fish gill cells. The excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of algae cells were DNA 376~398nm / 417~477nm, ER 442~502nm / 503~538nm, RNA491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Cy5588~668nm / 652~732nm, and the excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of fish gill cells were DNA 376~398nm / 417~477nm, ER442~502nm / 503~538nm, RNA 491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Mito 588~668nm / 672~712nm, 9 (3×3) imaging field points are set in each well of the well plate, 2×2 pixel merging is used, and each point automatically captures 5-color fluorescence channel images and 3 bright field channel images from different z-axis focuses; the automatically captured images are used to extract cell morphological features, and the cells, cell nuclei, and cytoplasm of each image are located. The image quality must meet the following conditions: the average image intensity is 10~240, the image focus score value is >0.5, the image edge front standard deviation is <0.2, the cell area is 50~500, the number of cell debris holes is <5, the cell density is >50, and the clarity of cell nuclear staining is >1.5. The gray-level co-occurrence matrix algorithm of texture feature analysis was used to calculate the morphology, intensity, texture, brightness, average grayscale of each cell and the minimum distance, adjacency value, and aggregation degree between each cell, and the 5797 cell morphological feature items of each cell and the arithmetic mean of the cell morphological feature value corresponding to each cell morphological feature item were obtained.

[0044] Step 3: Exclude the collinear crossing feature items within the phenotypic characteristic data of algae cells and fish gill cells, and retain the feature items whose feature values ​​are not equal to 0; use the Z-Score method and the maximum-minimum method to standardize the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering; classify and integrate the feature items according to the subcellular structures corresponding to the feature items, and arrange and classify the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering and feature value standardization into ultra-high dimensional feature items consisting of algae cell DNA, algae cell endoplasmic reticulum, algae cell nucleosomes and cytoplasmic RNA, algae cell actin and Golgi apparatus and plasma membrane, algae cell chloroplasts, algae cell bright field, fish gill cell DNA, fish gill cell endoplasmic reticulum, fish gill cell nucleosomes and cytoplasmic RNA, fish gill cell actin and Golgi apparatus and plasma membrane, fish gill cell mitochondria, and fish gill cell bright field to construct a toxicity matrix.

[0045] Step 4: Based on the acute toxicity effect values ​​and phenotypic characteristic data of algal cells and the acute toxicity effect values ​​and phenotypic characteristic data of fish gill cells, a machine learning model was established through a random forest model; the constructed toxicity matrix was subjected to dimensionality reduction processing using the least partial squares discriminant analysis method to obtain 12 water quality comprehensive toxicity characteristic variables, which were substituted into the machine learning model to obtain the comprehensive water quality toxicity of the full-process wastewater samples of Plant B, expressed as the concentration that can cause 10% maximum effect (EC10).

[0046] The subcellular structural images of algae cells and gill cells of the full-process samples of Plant B were obtained by this method. The cell morphological characteristics were extracted, the corresponding cell phenotypic characteristic data were obtained, and the toxicity matrix was constructed. The toxicity matrix of the full-process sewage samples of Plant B was subjected to feature item dimensionality reduction processing using the least partial squares discriminant analysis method. Each sample obtained 12 water quality comprehensive toxicity characteristic variables, which were brought into the machine learning model to obtain the comprehensive water quality toxicity of the sewage samples of Plant B's water inlet, aerated grit chamber, anoxic tank, aerobic tank, secondary sedimentation tank, sand filter, disinfection tank, and outlet, such as Figure 4 As shown, they are 36.0%, 42.3%, 67.8%, 56.3%, 58.3%, 64.4%, 56.2% and 60.6% respectively.

[0047] Example 3

[0048] The difference from Example 1 is that the application object of this example is the effluent samples of three municipal sewage treatment plants in the Beijing-Tianjin-Hebei region. The daily processing capacity of plants C, D, and E is 1.2 to 2.8 million cubic meters per day, and the effluent contains 42.00 to 58.89 mg / L of COD, 6.26 to 10.09 mg / L of total nitrogen, and 0.09 to 0.35 mg / L of total phosphorus. A high-throughput determination method for comprehensive toxicity of water quality, the specific establishment steps are:

[0049] Step 1: Filter the wastewater samples from plants C, D, and E through a 0.22 μm water filter membrane.

[0050] Step 2: Using the cells of the ceratospermum elata and the gill cells of rainbow trout as the test organisms, the filtered sewage samples were used to expose the cells of the ceratospermum elata and the gill cells of rainbow trout for 24 hours; Hoechst 33342 dye, concanavalinA / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye were used to prepare multiple fluorescent stains, and the exposed algal cells were subjected to multiple fluorescent staining, using Hoechst 33342 dye, concanavalinA / AlexaFluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye, MitoTracker Deep Red dye was used to make multiple fluorescent stains, and multiple fluorescent staining was performed on the gill cells after exposure to poison; a high-content cell imaging and analysis system was used to automatically collect subcellular structure images of algae cells and gill cells inoculated in 8 parallel experiments on the well plate with high throughput: a 63x water-immersion objective was used to obtain subcellular structure images of algae cells, and a 20x water-immersion objective was used to obtain subcellular structure images of gill cells. The excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of algae cells were DNA 376~398nm / 417~477nm, ER 442~502nm / 503~538nm, RNA491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Cy5588~668nm / 652~732nm, and the excitation / emission wavelengths of the 5-color fluorescent channels used for automatic imaging of fish gill cells were DNA 376~398nm / 417~477nm, ER442~502nm / 503~538nm, RNA 491~571nm / 573~613nm, AGP 502~622nm / 622~662nm, Mito 588~668nm / 672~712nm, 9 (3×3) imaging field points are set in each well of the well plate, 2×2 pixel merging is used, and each point automatically captures 5-color fluorescence channel images and 3 bright field channel images from different z-axis focuses; the automatically captured images are used to extract cell morphological features, and the cells, cell nuclei, and cytoplasm of each image are located. The image quality must meet the following conditions: the average image intensity is 10~240, the image focus score value is >0.5, the image edge front standard deviation is <0.2, the cell area is 50~500, the number of cell debris holes is <5, the cell density is >50, and the clarity of cell nuclear staining is >1.5. The gray-level co-occurrence matrix algorithm of texture feature analysis was used to calculate the morphology, intensity, texture, brightness, average grayscale of each cell and the minimum distance, adjacency value, and aggregation degree between each cell, and the 5797 cell morphological feature items of each cell and the arithmetic mean of the cell morphological feature value corresponding to each cell morphological feature item were obtained.

[0051] Step 3: Exclude the collinear crossing feature items within the phenotypic characteristic data of algae cells and fish gill cells, and retain the feature items whose feature values ​​are not equal to 0; use the Z-Score method and the maximum-minimum method to standardize the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering; classify and integrate the feature items according to the subcellular structures corresponding to the feature items, and arrange and classify the phenotypic characteristic data of algae cells and fish gill cells after feature item filtering and feature value standardization into ultra-high dimensional feature items consisting of algae cell DNA, algae cell endoplasmic reticulum, algae cell nucleosomes and cytoplasmic RNA, algae cell actin and Golgi apparatus and plasma membrane, algae cell chloroplasts, algae cell bright field, fish gill cell DNA, fish gill cell endoplasmic reticulum, fish gill cell nucleosomes and cytoplasmic RNA, fish gill cell actin and Golgi apparatus and plasma membrane, fish gill cell mitochondria, and fish gill cell bright field to construct a toxicity matrix.

[0052] Step 4: Based on the acute toxicity effect values ​​and phenotypic characteristic data of algal cells and the acute toxicity effect values ​​and phenotypic characteristic data of fish gill cells, a machine learning model was established through a random forest model; the constructed toxicity matrix was subjected to dimensionality reduction processing using the least partial squares discriminant analysis method to obtain 12 water quality comprehensive toxicity characteristic variables, which were substituted into the machine learning model to obtain the comprehensive water quality toxicity of the effluent from plants C, D, and E, expressed as the concentration that can cause 10% of the maximum effect (EC10).

[0053] The subcellular structure images of algae cells and gill cells of the effluent from plants C, D, and E obtained by this method. The cell morphological characteristics were extracted to obtain the corresponding cell phenotypic characteristic data and construct a toxicity matrix. The toxicity matrix of the wastewater samples from plants C, D, and E was processed by the least partial squares discriminant analysis method to reduce the dimension of the feature items. Each sample obtained 12 water quality comprehensive toxicity characteristic variables, which were brought into the machine learning model to obtain the comprehensive water quality toxicity of the effluent from plants C, D, and E, as shown in Figure 2. Figure 5 As shown, they are 47.0%, 56.9% and 47.9% respectively.

Claims

1. A high-throughput method for determining comprehensive toxicity of water quality, characterized in that: The following steps are involved: After exposing the test organisms to the toxicants using sewage samples, the phenotypic characteristic data of the test organisms are obtained; Constructing a toxicity matrix; A machine learning model was established and combined with the toxicity matrix to determine the comprehensive toxicity of water quality of sewage samples.

2. The high-throughput determination method for comprehensive toxicity of water quality according to claim 1, characterized in that: Before using the sewage samples to expose the test organisms to toxic substances, the sewage samples were pretreated.

3. The high-throughput determination method for comprehensive toxicity of water quality according to claim 1, characterized in that: The test organisms are algae cells and fish gill cells.

4. The method for high-throughput determination of comprehensive toxicity of water quality according to claim 3, characterized in that: After the test organisms were exposed to the poison using sewage samples, the phenotypic characteristic data of the test organisms were obtained using multiple fluorescent staining, high-content automatic imaging and cell morphological feature extraction.

5. The high-throughput determination method for comprehensive toxicity of water quality according to claim 4, characterized in that: The multiplex fluorescent stain used for algal cell staining consists of Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye; the multiplex fluorescent stain used for fish gill cell staining consists of Hoechst 33342 dye, concanavalinA / AlexaFluor 488 dye, SYTO 14 dye, phalloidin / AlexaFluor 568 and wheat-germ agglutinin / Alexa Fluor 555 dye, and MitoTracker Deep Red dye.

6. The high-throughput determination method for comprehensive toxicity of water quality according to claim 4, characterized in that: The high-content automatic imaging uses a high-content cell imaging and analysis system to automatically collect subcellular structure images of algae cells and fish gill cells inoculated in 4 to 8 groups of parallel experiments on a well plate with high throughput.

7. The high-throughput determination method for comprehensive toxicity of water quality according to claim 4, characterized in that: The steps of extracting cell morphological features are as follows: positioning the cells, cell nuclei and cytoplasm of each image, and the image quality must meet the following conditions: the average image intensity is 10-240, the image focus score is >0.5, the image edge pre-degree standard deviation is <0.2, the cell area is 50-500, the number of cell debris holes is <5, the cell density is >50, and the cell nuclear staining clarity is >1.5; the gray level co-occurrence matrix algorithm of texture feature analysis is used to calculate the morphology, intensity, texture, brightness, average grayscale of each cell and the minimum distance, adjacency value and aggregation degree between each cell, and obtain the cell morphological feature items of each cell and the arithmetic mean of the cell morphological feature value corresponding to each cell morphological feature item.

8. The high-throughput determination method for comprehensive toxicity of water quality according to claim 3, characterized in that: The toxicity matrix is ​​composed of the phenotypic characteristic data of algae cells and fish gill cells, which are clustered and arranged after feature item filtering and feature value standardization; the feature item filtering is to exclude the feature items with collinear intersections and retain the feature items with eigenvalues ​​not equal to 0; the feature value standardization method is the Z-Score method and the maximum-minimum method; the cluster arrangement method is to classify and integrate the feature items according to the subcellular structures corresponding to the feature items, and the arrangement classification is ultra-high dimensional feature items consisting of algae cell DNA, algae cell endoplasmic reticulum, algae cell nucleosomes and cytoplasmic RNA, algae cell actin and Golgi body and plasma membrane, algae cell chloroplasts, algae cell bright field, fish gill cell DNA, fish gill cell endoplasmic reticulum, fish gill cell nucleosomes and cytoplasmic RNA, fish gill cell actin and Golgi body and plasma membrane, fish gill cell mitochondria, and fish gill cell bright field.

9. The high-throughput determination method for comprehensive toxicity of water quality according to claim 3, characterized in that: The machine learning model is based on the acute toxicity effect value and phenotypic characteristic data of algal cells and the acute toxicity effect value and phenotypic characteristic data of fish gill cells, and is established through one of a random forest model, an XGBoost algorithm model, a Lasso regression algorithm model, a content-based recommendation algorithm model, and a support vector machine model.

10. The high-throughput determination method for comprehensive toxicity of water quality according to claim 1, characterized in that: The method for determining the comprehensive toxicity of water quality of sewage samples is as follows: using the least partial squares discriminant analysis method to perform dimension reduction processing on the constructed toxicity matrix, obtaining the characteristic variables of comprehensive toxicity of water quality, substituting them into the machine learning model to obtain the comprehensive toxicity of water quality of sewage samples; the comprehensive toxicity of water quality of sewage samples is the value of the acute toxic effect caused by the sewage samples, with the concentration that can cause 10% of the maximum effect (EC 10 )express.

Images