A method for measuring overall water toxicity using high throughput.

Abstract

This provides a high-throughput method for determining overall water quality toxicity. [Solution] The present invention discloses a high-throughput method for determining the overall toxicity of water quality, the method comprising: exposing target organisms to toxicity using wastewater samples, obtaining phenotypic characteristic data of the target organisms, constructing a toxicity matrix, and establishing a machine learning model for determining the overall toxicity of wastewater samples in combination with the toxicity matrix. The method for determining overall water quality toxicity provided by the present invention does not require sample concentration or enrichment, guarantees the comprehensiveness and accuracy of the test results, and can be applied to high-throughput determination of the overall toxicity of the entire wastewater process from wastewater treatment plants.

Description

Technical Field

[0001] The present invention belongs to the field of water quality risk management and relates to a method for measuring the overall toxicity of water quality with high throughput.

Background Art

[0002] The wastewater generated in the human life and production processes has complex components, large fluctuations in water quality, and contains various pollutants, so it has become one of the main factors causing water ecological safety problems. With the progress of industrialization and urbanization, the pollutants generated from human life and production activities are increasing, and new pollutants with extremely low concentrations but high risks (such as antibiotics, microplastics and their derivatives, perfluorinated and polyfluorinated alkyl substances, brominated flame retardants, etc.) enter the wastewater, significantly increasing the potential risk of wastewater, and the global concern about the toxicity of wastewater is increasing.

[0003] Due to the variety of pollutants contained in wastewater, pollutants not only have a direct toxic effect on organisms, but may also combine with other pollutants to form a complex toxic effect. Therefore, it is difficult for the concentration level or toxicity effect value of a single or specific pollutant to accurately reflect the overall toxicity effect of wastewater, which may cause significant deviations. The general methods for evaluating the overall toxicity effect of wastewater used at home and abroad include the wastewater comprehensive toxicity method, the toxicity identification and evaluation method, the direct toxicity evaluation method, etc. Among them, the wastewater comprehensive toxicity evaluation method is a method of exposing a sample in which a standard model organism is diluted step by step and measuring the acute toxicity value given by the sample to the test organism during a fixed exposure time.

[0004] However, these methods require a significant amount of time for numerous stepwise dilution sample exposure experiments to determine acute toxicity values, making it difficult to rapidly detect the toxic effects of large quantities of wastewater samples in a short time. Furthermore, there are differences in the sensitivity and tolerance of test organisms to different pollutants and water quality, and these differences are particularly large in wastewater from different treatment processes. Therefore, there is an urgent need for a highly sensitive and adaptable high-throughput comprehensive toxicity measurement method that can be applied to the entire process of wastewater treatment plants. [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] The object of the present invention is to provide a high-throughput method for determining overall water toxicity in order to improve upon existing methods for evaluating the overall toxic effects of wastewater, which have low detection throughput, variability in the sensitivity and tolerance of target organisms used in toxicity testing for different pollutants / water quality, and large variability in wastewater across different treatment process segments. [Means for solving the problem]

[0006] The method for determining the overall toxicity of water quality in high throughput as described in the present invention comprises the following steps: acquiring phenotypic characteristic data of a target organism after exposure to toxicity using a sewage sample; constructing a toxicity matrix; and establishing a machine learning model for determining the overall toxicity of a sewage sample in combination with the toxicity matrix.

[0007] Preferably, the exposure time to contamination is 24 hours.

[0008] Preferably, the wastewater sample is pre-treated before use in order to expose the target organism to contamination.

[0009] Preferably, the pretreatment step for the waste liquid sample is to filter the waste liquid sample through a 0.22 μm aqueous filter membrane.

[0010] Preferably, the target organisms are algal cells and gill cells.

[0011] More preferably, the target organisms are lambda lunata cells and rainbow trout gill cells.

[0012] Preferably, after exposure staining of the target organism using the leaked sample, phenotypic characteristic data of the target organism is obtained using multiple fluorescent staining methods, high-content automated imaging, and cell morphological feature extraction.

[0013] Preferably, the multiplex fluorescent stain for staining algal cells 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 for staining gill cells consists of Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 containing wheat germ agglutinin / Alexa Fluor 555 dye, Fluor 555 dye, and MitoTracker Deep Red dye.

[0014] Preferably, the high-content automated imaging involves the use of a high-content cell imaging and analysis system to automatically acquire intracellular structural images of algal cells and gill cells inoculated into well plates at high throughput in 4 to 8 sets of parallel experiments.

[0015] Preferably, the image acquisition conditions for the high-concentration cell imaging and analysis system are as follows: 9 (3-3) imaging field spots are set in each well of the well plate, 2-2 pixel merging is used, and 5-color fluorescence channel images are automatically acquired at each spot along with 3 bright-field channel images originating from different z-axis foci; intracellular structure images of algal cells are acquired using a 63x immersion objective lens, and intracellular structure images of gill cells are acquired using a 20x immersion objective lens. The excitation / emission wavelengths of the 5-color fluorescence channels for automated imaging of algal cells were 376-398 nm / 417-477 nm for DNA, 442-502 nm / 503-538 nm for ER, 491-571 nm / 573-613 nm for RNA, 502-622 nm / 622-662 nm for AGP, and 588-666 nm for Cy5. The excitation / emission wavelengths for the five-color fluorescence channels for automated gill cell imaging are: DNA 376-398 nm / 417-477 nm, ER 442-502 nm / 503-538 nm, RNA 491-571 nm / 573-613 nm, AGP 502-622 nm / 622-662 nm / 622-662 nm, and Mito 588-668 nm / 672-712 nm.

[0016] Preferably, the steps for extracting cell morphological features are as follows: Identify the positions of cells, nuclei, and cytoplasm in each image, and the image quality should meet the following conditions: the average image intensity is 10-240, the image focus score is 0.5 or higher, the standard deviation of the frontality of the image edges is less than 0.2, the cell area is 50-500, and the number of holes in cell fragments is less than 5. Cell area is 50-500, cell density is 50 or higher, and the clarity of cell nucleus staining is 1.5 or higher; Using the grayscale covariance matrix algorithm for texture feature analysis, calculate morphology, intensity, texture, brightness, mean grayscale, minimum distance between cells, neighbor value, and cell cohesion, and obtain the cell morphological feature items for each cell and the arithmetic values ​​of the corresponding cell morphological feature values ​​for each cell morphological feature item. Obtain the average value for each cell and the arithmetic values ​​of the cell morphological features corresponding to each cell morphological feature.

[0017] Preferably, 5797 cellular morphological features were obtained for each cell.

[0018] Preferably, the toxicity matrix consists of phenotypic characteristic data of algal cells and gill cells after feature filtering, eigenvalue standardization, and clustering arrangement, the feature filtering involves excluding eigenvalues ​​that intersect at a common line and retaining eigenvalues ​​that are not equal to 0, the eigenvalue standardization method is the Z-Score method and the maximum-minimum method, and the clustering arrangement method involves classifying and integrating eigenvalues ​​according to the small cell structure corresponding to the feature items. The classification of the feature items is performed based on the intracellular structure corresponding to the feature items, and is a super-high-dimensional feature consisting of bright-field images of fish gill cells, comprising algal cell DNA, algal cell endoplasmic reticulum, algal cell nucleosomes and cytoplasmic RNA, algal cell actin having Golgi apparatus and cell membrane, algal cell chloroplasts, algal cell bright-field images, gill cell DNA, gill cell endoplasmic reticulum, gill cell nucleosomes and cytoplasmic RNA, gill cell actin having Golgi apparatus and cell membrane, and gill cell mitochondria.

[0019] Preferably, the machine learning model is based on acute toxicity effect values ​​and phenotypic characteristics data of algal cells and acute toxicity effect values ​​and phenotypic characteristics data of gill cells, and is constructed using one of the following: 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.

[0020] Preferably, the method for obtaining the integrated water quality toxicity of the drainage sample is as follows: perform dimensionality reduction processing on the constructed toxicity matrix using the least squares discriminant analysis method to obtain integrated water quality toxicity characteristic variables, substitute them into a machine learning model to obtain the integrated toxicity of the drainage sample; the integrated toxicity of the drainage sample is the value of the acute toxicity effect caused by the drainage sample, and is represented by the concentration (EC10) that can cause 10% of the maximum effect. The water quality toxicity of the sewage sample is the value of the acute toxicity effect caused by the sewage sample, and is represented by the concentration (EC10) that can cause 10% of the maximum effect.

[0021] Preferably, the number of the water quality composite toxicity characteristic variables is 12.

[0022] Advantageous effects: Compared with the prior art, the present invention has the following important advantages: (1) The present invention provides a high-throughput and rapid comprehensive toxicity detection method for sewage water quality, which solves the problems of complicated and time-consuming detection steps and low detection throughput in the existing methods.

[0023] (2) Based on the high-annotation cell imaging technology and the machine learning model, the present invention automatically grasps the comprehensive toxicity effect of the sewage sample at the cell and subcellular structure levels, and at the same time, accurately quantifies the acute toxicity effect of the sewage sample, greatly simplifies the measurement steps, and ensures the comprehensiveness and accuracy of the test results.

[0024] (3) By constructing a toxicity matrix using the different response characteristics of algal cells and gill cells to the toxic effect of sewage samples, the present invention reduces the selectivity of the existing method for specific sewage samples, improves the universality of the high-throughput method for measuring the comprehensive toxicity of water for all types of sewage samples, and improves the practicality of the method.

Brief Description of the Drawings

[0025] [Figure 1] It is a flowchart of the method of the present invention for measuring composite water quality toxicity with high throughput; [Figure 2]An intracellular structure image of algal cells and gill cells in Example 1 of the present invention; [Figure 3] Shows a toxicity matrix constructed from phenotypic characteristic data of algae and gill cells in Example 1 of the present invention; [Figure 4] Shows the combined water quality toxicity of the total process wastewater sample from Plant B in Example 2 of the present invention; [Figure 5] Shows the combined toxicity of the water quality of the wastewater from Plants C, D, and E in Example 3 of the present invention.

Mode for Carrying Out the Invention

[0026] The technical solution of the present invention will be further described below in connection with embodiments and the accompanying drawings.

Examples

[0027] This example is a municipal wastewater treatment plant in Jiangsu Province. The daily treatment capacity of Factory A is 80,000 m 3 / day, the influent COD is 254.0 mg / L, the total nitrogen is 29.27 mg / L, and the total phosphorus is 2.07 mg / L:

[0028] Step 1: The wastewater sample from Factory A was filtered through a 0.22 μm aqueous filter membrane.

[0029] Step 2: Hornwort cells and rainbow trout gill cells were used as test organisms. Filtered sewage samples were exposed to and infected with hornwort cells and rainbow trout gill cells for 24 hours. Algal cells were stained with Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568, and wheat germ agglutinin / Alexa Fluor 555 dyes, followed by multiplex fluorescence staining. Fish gill cells were stained with Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568, and wheat germ agglutinin / Alexa Fluor 555 dyes, followed by multiplex fluorescence staining. Multichannel fluorescent staining was performed using Alexa Fluor 555 and MitoTracker Deep Red fluorescent dyes, followed by multichannel fluorescent staining of fish gill cells. A high-content cell imaging and analysis system was used to automatically collect images of the intracellular structure of algal and gill cells seeded in 6-well plates in six parallel experiments with high processing capacity: intracellular structure of algal cells was acquired using a 63× immersion objective lens, and intracellular structure of gill cells was acquired using a 20× immersion objective lens. The excitation / emission wavelengths of the five fluorescence channels used for automated image acquisition of algal cells were: DNA 376~398 nm / 417~477 nm, ER 442~502 nm / 503~538 nm, RNA 491~571 nm / 573~613 nm, AGP 502~622 nm / 622~662 nm, Cy5 588~668 nm / 652~732 nm. The excitation / emission wavelengths of the five fluorescence channels used for automated imaging of gill cells were: DNA 376~398 nm / 417~477 nm, ER 442~502 nm / 503~538 nm, RNA 491~571 nm / 573~613 nm, AGP 502~622 nm / 622~662 nm, Mito 588~668 nm / 672~712 nm. Nine (3x3) imaging field points are set up in each well of the well plate, and 2x2 pixel binning is used.Each point consists of three points selected from five fluorescence channels and three different z-axis focus points. Automatically captured images are used to extract morphological features of cells, identifying the location of cells, nuclei, and cytoplasm in each image. Image quality must meet the following conditions: average image intensity of 10-240, image focus score of 0.5 or higher, standard deviation of image edges less than 0.2, cell area of ​​50-500, number of holes in cell fragments of 5, cell density of 50 or higher, and nuclear stain clarity of 1.5 or higher. Using a grayscale co-occurrence matrix algorithm with texture feature analysis, the shape, intensity, texture, brightness, average grayscale level, minimum distance between cells, proximity value, and degree of clustering of each cell are calculated, and the 5797 cellular morphological feature items and the arithmetic mean of the cellular morphological feature values ​​corresponding to each cellular morphological feature item are determined.

[0030] Step 3: Remove collinear features from the phenotypic feature data of algal cells and gill cells, and retain features whose feature values ​​are not equal to 0. Standardize the filtered phenotypic feature data of algal cells and gill cells using the Z-Score method and the minimum-maximum method. Classify and integrate features based on the intracellular structure corresponding to the feature item. After filtering and standardization, organize and classify the phenotypic feature data of algal cells and gill cells into the following categories. Construct a toxicity matrix using ultra-high-dimensional features of algal cell DNA, algal cell intraplasmic network, algal cell nucleosomes and cytoplasmic RNA, algal cell actin, Golgi apparatus and cell membrane, algal cell chloroplasts, algal cell bright-field, gill cell DNA, gill cell intraplasmic network, gill cell nucleosomes and cytoplasmic RNA, gill cell actin, Golgi apparatus and cell membrane, gill cell mitochondria, and gill cell bright-field.

[0031] Step 4: Based on the acute toxicity effect values ​​and phenotypic feature data of algal cells and gill cells, a machine learning model is constructed using a random forest model. Furthermore, dimensionality reduction is performed on the feature terms of the constructed toxicity matrix using least-squares discriminant analysis to obtain 12 overall water quality toxicity feature variables. These results are then substituted into the machine learning model to evaluate the overall water quality toxicity of the influent water of Plant A. This toxicity is expressed as the concentration (EC10) that causes 10% of the maximum effect.

[0032] As shown in Figure 2, intracellular structural images of algal cells and gill cells from the inflow sample from Plant A obtained using this method are shown. As shown in Figure 3, cellular morphological features were extracted from these images, and cellular phenotypic feature data of algal cells and gill cells were obtained to construct a toxicity matrix. As shown in Table 1, dimensionality reduction processing of the feature terms of the toxicity matrix was performed using least-squares discriminant analysis to obtain 12 overall water quality toxicity characteristic variables. When these were introduced into a machine learning model, the overall water quality toxicity of the inflow water from Plant A was found to be 55.2%.

[0033] The 12 water quality overall toxicity characteristic variables obtained in Example 1

[0034] [Table 1] [Examples]

[0035] Unlike Example 1, this example applies to sewage samples from the entire process of a municipal wastewater treatment plant located in the southwest region, including wastewater samples from the intake, aeration and sedimentation tank, anoxic tank, aerobic tank, secondary sedimentation tank, sand filtration tank, disinfection tank, and discharge outlet. Plant B has a daily processing capacity of 450,000 m³. 3 The COD at the intake was 241.1 mg / L, total nitrogen was 27.02 mg / L, and total phosphorus was 2.94 mg / L; the COD at the discharge was 55.40 mg / L, total nitrogen was 10.37 mg / L, and total phosphorus was 0.38 mg / L. A high-throughput method for measuring overall water toxicity was established using the following steps:

[0036] Step 1: Eight wastewater samples obtained from the entire process at Plant B are filtered using a 0.22 μm aqueous filter membrane.

[0037] Step 2: Sheephead lunar alga cells and rainbow trout gill cells are used as test organisms. Sheephead lunar alga cells and rainbow trout gill cells are exposed to a staining toxin for 24 hours using filtered wastewater samples. Multiple fluorescence staining is performed on the exposed algal cells using Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat germ agglutinin / Alexa Fluor 555. Furthermore, multiple fluorescence staining is performed on the exposed gill cells using Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568, wheat germ agglutinin / Alexa Fluor 555 dye and MitoTracker Deep Red dye. Using a high-throughput intracellular imaging and analysis system, intracellular structural images of algal and gill cells inoculated into perforated plates for four parallel experiments are automatically acquired. Intracellular structural images of algal cells are acquired using a 63x immersion objective lens, and intracellular structural images of gill cells are acquired using a 20x immersion objective lens. The excitation / emission wavelengths of the five fluorescence channels used for automated imaging of algal cells are as follows: DNA 376398 nm / 417477 nm, ER 442502 nm / 503538 nm, RNA 491571 nm / 573613 nm, AGP 502622 nm / 622662 nm, Cy5 588668 nm / 652732 nm. The excitation / emission wavelengths of the five fluorescence channels used for automated imaging of gill cells are 376398 nm / 417477 nm for DNA, 442502 nm / 503538 nm for ER, 491571 nm / 573613 nm for RNA, 502622 nm / 622662 nm for AGP, and 588668 nm / 672712 nm for Mito. Nine (3x3) imaging field points are set up in each well of the perforated plate, using 2x2 pixel binning. Each point automatically acquires a five-color fluorescence channel image and three bright-field channel images from different z-axis focuses. The automatically acquired images are used for cellular morphological feature extraction, identifying the location of cells, nuclei, and cytoplasm in each image.The image quality must meet the following conditions: mean image intensity of 10-240, image focus score of 0.5 or higher, standard deviation of image edges less than 0.2, cell area of ​​50-500, number of holes in cell debris less than 5, cell density of 50 or higher, and cell nucleus staining clarity of 1.5 or higher. For texture feature analysis, a grayscale co-occurrence matrix algorithm is used to calculate the morphology, intensity, texture, brightness, mean grayscale, minimum distance between cells, proximity value, and cohesion for each cell. The 5797 cellular morphological feature items for each cell and the arithmetic mean of the cellular morphological feature values ​​corresponding to each cellular morphological feature item are then calculated.

[0038] Step 3: Remove collinear features from the phenotypic feature data of algal cells and gill cells, and retain features whose feature values ​​are not equal to 0. Standardize the filtered phenotypic feature data of algal cells and gill cells using the Z-Score method and the maximum-minimum method. After classifying and integrating features based on the intracellular structure corresponding to the feature item, filtering and standardization are performed, and the phenotypic feature data of algal cells and gill cells are organized and classified into the following categories: algal cell DNA, algal cell cellular network, algal cell nucleosomes and cytoplasmic RNA, algal cell actin, Golgi apparatus and cell membrane, algal cell chloroplasts, algal cell bright-field view, gill cell DNA, gill cell cellular network, gill cell nucleosomes and cytoplasmic RNA, gill cell actin, Golgi apparatus and cell membrane, gill cell mitochondria, gill cell bright-field view. These are the ultra-high-dimensional features that construct the toxic matrix.

[0039] Step 4: Based on the acute toxicity effect values ​​and phenotypic feature data of algal cells and gill cells, a machine learning model is constructed using a random forest model. The dimensionality of the feature terms of the constructed toxicity matrix is ​​reduced using least-partial-squares discriminant analysis to obtain 12 overall water quality toxicity feature variables, which are then introduced into the machine learning model to evaluate the overall water quality toxicity of wastewater samples from all processes of Plant B. This toxicity is expressed as the concentration (EC10) that causes 10% of the maximum effect.

[0040] The intracellular structure images of algal cells and gill cells from all process samples obtained by this method are shown. Morphological features of the cells were extracted from these images, and corresponding phenotypic feature data were obtained to construct a toxicity matrix. The dimensionality of the feature terms in the toxicity matrix of wastewater samples from all processes of Plant B was reduced using least-squares discriminant analysis, and 12 overall water quality toxicity feature variables were obtained from each sample and introduced into a machine learning model. As shown in Figure 4, the overall water quality toxicity of wastewater samples from the intake, aeration and sedimentation tank, anaerobic tank, aerobic tank, secondary sedimentation tank, sand filtration tank, disinfection tank, and discharge outlet of Plant B were 36.0%, 42.3%, 67.8%, 56.3%, 58.3%, 64.4%, 56.2%, and 60.6%, respectively. [Examples]

[0041] Unlike Example 1, this example applies to wastewater samples from three urban wastewater treatment plants in the Beijing-Tianjin-Hebei region. The daily treatment capacity of plants C, D, and E is 1.2 million to 2.8 million m³. 3 The wastewater contains COD of 42.00–58.89 mg / L, total nitrogen of 6.26–10.09 mg / L, and total phosphorus of 0.09–0.35 mg / L. A high-throughput method for measuring overall water toxicity was established in the following steps:

[0042] Step 1: Filter wastewater samples from plants C, D, and E through a 0.22 μm aqueous filter membrane.

[0043] Step 2: Sheephead lunar alga cells and rainbow trout gill cells are used as test organisms. Sheephead lunar alga cells and rainbow trout gill cells are exposed to a staining toxin for 24 hours using filtered wastewater samples. Algal cells after exposure are stained with multiple fluorescence using Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568 and wheat germ agglutinin / Alexa Fluor 555. Furthermore, gill cells after exposure are also stained with multiple fluorescence using Hoechst 33342 dye, concanavalin A / Alexa Fluor 488 dye, SYTO 14 dye, phalloidin / Alexa Fluor 568, wheat germ agglutinin / Alexa Fluor 555 and MitoTracker Deep Red dye. Using a high-intensity cell imaging and analysis system, eight parallel experiments were conducted to automatically collect intracellular structural images of inoculated algal and gill cells at high throughput. Intracellular structural images of algal cells were acquired using a 63x water immersion objective lens, and intracellular structural images of gill cells were acquired using a 20x water immersion objective lens. The excitation / emission wavelengths of the five fluorescence channels used for automated imaging of algal cells were DNA 376~398 nm / 417~477 nm, ER 442~502 nm / 503~538 nm, RNA 491~571 nm / 573~613 nm, AGP 502~622 nm / 622~662 nm, and Cy5 588~668 nm / 652~732 nm. The excitation / emission wavelengths of the five fluorescence channels used for automated imaging of gill cells are: DNA 376-398 nm / 417-477 nm, ER 442-502 nm / 503-538 nm, RNA 491-571 nm / 573-613 nm, AGP 502-622 nm / 622-662 nm, and Mito 588-668 nm / 672-712 nm. Nine (3x3) imaging field points are set up in each well of the perforated plate, and using 2x2 pixel binning, each point automatically acquires a five-color fluorescence channel image and three bright-field channel images from different z-axis focuses. Cellular morphological features are extracted using the automatically acquired images to identify the location of cells, nuclei, and cytoplasm in each image.The image quality must meet the following conditions: mean image intensity of 10-240, image focus score of 0.5 or higher, standard deviation of image edges less than 0.2, cell area of ​​50-500, number of holes in cell debris less than 5, cell density of 50 or higher, and cell nucleus staining clarity of 1.5 or higher. For texture feature analysis, the grayscale co-occurrence matrix algorithm is used to calculate the morphology, intensity, texture, brightness, mean grayscale, minimum distance between cells, proximity value, and cohesion of each cell. The 5797 cellular morphological feature items for each cell and the arithmetic mean of the cellular morphological feature values ​​corresponding to those morphological feature items are then calculated.

[0044] Step 3: Remove collinear features from the phenotypic feature data of algal cells and gill cells, and retain features whose feature values ​​are not equal to 0. Standardize the filtered phenotypic feature data of algal cells and gill cells using the Z-Score method and the maximum-minimum method. After classifying and integrating features based on the intracellular structure corresponding to the feature item, and after filtering and standardization, organize and classify the phenotypic feature data of algal cells and gill cells into the following categories: Algal cell DNA, Algal cell intraplasmic network, Algal cell nucleosomes and cytoplasmic RNA, Algal cell actin, Golgi apparatus and cell membrane, Algal cell chloroplasts, Algal cell bright-field, Gill cell DNA, Gill cell intraplasmic network, Gill cell nucleosomes and cytoplasmic RNA, Gill cell actin, Golgi apparatus and cell membrane, Gill cell mitochondria, and Gill cell bright-field. Construct a toxicity matrix using these ultra-high-dimensional features.

[0045] Step 4: Based on the acute toxicity effect values ​​and phenotypic feature data of algal cells and gill cells, a machine learning model is constructed using a random forest model. The dimensionality of the feature terms of the constructed toxicity matrix is ​​reduced using the least-partial-squares discriminant analysis method to obtain 12 water quality overall toxicity feature variables, which are then introduced into the machine learning model. This allows for the evaluation of the overall water quality toxicity of wastewater from plants C, D, and E, and this toxicity is expressed as the concentration (EC10) that causes 10% of the maximum effect.

[0046] The intracellular structure images of algal cells and gill cells contained in the wastewater from plants C, D, and E obtained by this method are shown. Morphological features of the cells were extracted from these images, and phenotypic feature data of the corresponding cells were obtained to construct a toxicity matrix. Using least-squares discriminant analysis, the dimensionality of the feature terms in the toxicity matrix of wastewater samples from plants C, D, and E was reduced, and 12 overall water quality toxicity feature variables were obtained from each sample and introduced into a machine learning model. As a result, the overall water quality toxicity of the wastewater from plants C, D, and E was 47.0%, 56.9%, and 47.9%, respectively, as shown in Figure 5.

Claims

1. A high-throughput method for measuring overall water toxicity, comprising the following steps: The test organism is exposed to toxicity using wastewater samples, and tabular characteristic data of the test organism is obtained; Construct a toxicity matrix; We will build a machine learning model and measure the overall water toxicity of wastewater samples using a toxicity matrix; After exposing the test organisms to toxicity using wastewater samples, tabular characteristic data of the test organisms is obtained using multifluorescence staining, high-content automated imaging, and extraction of cellular morphological features; The aforementioned high-content automated imaging system is characterized by using a high-content cell imaging and analysis system to automatically collect intracellular structural images of algal cells and gill cells inoculated into 4 to 8 parallel experiments at high throughput; The organisms under test are characterized by being algal cells and gill cells; The toxicity matrix is ​​characterized by being constructed by filtering feature terms and standardizing feature values ​​based on phenotypic characteristic data of algal cells and gill cells, followed by clustering; The filtering of the feature terms is characterized by excluding collinear feature terms and retaining feature terms whose feature values ​​are not equal to 0; The standardization method for the aforementioned feature values ​​is characterized by being the Z-Score method and the maximum-minimum method; The clustering and alignment method is characterized by classifying and integrating feature terms based on intracellular structures corresponding to the feature terms, and being composed of ultra-high-dimensional feature terms consisting of algal cell DNA, algal cell cellular network, algal cell nucleosomes and cytoplasmic RNA, algal cell actin, Golgi apparatus and cell membrane, algal cell chloroplasts, algal cell bright-field images, gill cell DNA, gill cell cellular network, gill cell nucleosomes and cytoplasmic RNA, gill cell actin, Golgi apparatus and cell membrane, gill cell mitochondria, and gill cell bright-field images.

2. A method for measuring the overall toxicity of water quality according to claim 1 in high throughput, characterized in that the wastewater sample is pretreated before the test organism is exposed to the toxicity.

3. A method for measuring the overall toxicity of water quality according to claim 1 in high throughput, characterized in that the multifluorescent stain used for staining algal cells is composed 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, and the multifluorescent stain used for staining gill cells is composed 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 and MitoTracker Deep Red dye.

4. A method for measuring the overall toxicity of water quality in high throughput according to claim 1, wherein the step of extracting cellular morphological features includes identifying the location of cells, nuclei, and cytoplasm in each image, and the image quality must satisfy the following conditions: mean image intensity of 10 to 240, image focus score greater than 0.5, standard deviation of image edges less than 0.2, cell area of ​​50 to 500, number of holes in cell debris less than 5, cell density of 50 or more, and cell nucleus staining clarity of 1.5 or more; a grayscale co-occurrence matrix algorithm is used for texture feature analysis to calculate the morphology, intensity, texture, brightness, mean grayscale, minimum distance between cells, proximity value, and degree of aggregation of each cell, and the arithmetic mean of the cellular morphological feature items and the cellular morphological feature values ​​corresponding to each cellular morphological feature item is obtained.

5. A method for measuring the overall toxicity of water quality according to claim 1 in high throughput, characterized in that the machine learning model is constructed using one of the following: a random forest model, an XGBoost algorithm model, a Lasso regression algorithm model, a content-based recommendation algorithm model, or a support vector machine model, based on acute toxicity effect values ​​and phenotypic characteristic data of algal cells and acute toxicity effect values ​​and phenotypic characteristic data of gill cells.

6. A method for measuring the overall water toxicity described in claim 1 in high throughput, wherein the method for measuring the overall water toxicity of a wastewater sample is characterized by reducing the dimensionality of feature terms using least-squares discriminant analysis on a constructed toxicity matrix to obtain overall water toxicity feature variables, and then introducing the results into a machine learning model to obtain the overall water toxicity of a wastewater sample; The overall toxicity of the wastewater sample is a numerical value representing the acute toxic effect caused by the wastewater sample, and is expressed as the concentration (EC10) that causes 10% of the maximum effect.

Images