Tetracycline photocatalytic degradation process toxicity prediction method based on machine learning

Abstract

The invention relates to a tetracycline photocatalytic degradation process toxicity prediction method based on machine learning. The tetracycline photocatalytic degradation process toxicity prediction method comprises the following steps: determining an intermediate product in a tetracycline photocatalytic degradation process; speculating a reaction path in the process of photocatalytic degradation of tetracycline; simulating the acute toxicity of an intermediate product to aquatic organisms, and quantifying by using a semi-lethal concentration; establishing a machine learning model; inputting to-be-predicted data into the machine learning model for prediction; toxicity prediction is carried out by combining the model toxicity corresponding to different types of simulated intermediate products and the predicted number of different intermediate products. The method can significantly reduce toxicology test verification experiments in toxicity simulation evaluation, and is simple and easy to implement; on the basis of a machine learning model, the content of the intermediate product under different reaction conditions can be simulated and predicted, the intermediate product is combined with toxicity simulation for evaluation, an experimenter can select a reaction stage with low toxicity under different conditions, resources and treatment time are saved, and treatment efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of environmental science photocatalytic degradation of tetracycline, and in particular to a method for predicting toxicity of a tetracycline photocatalytic degradation process based on machine learning. Background Art

[0002] Tetracycline is a broad-spectrum antibiotic that plays an important role in medicine and aquaculture. It is seriously polluted in the water environment, which is prone to produce drug-resistant bacteria and has an impact on human health. When photocatalytically treating wastewater containing tetracycline, reducing biological toxicity is one of the important goals. Traditional toxicity assessment often requires time-consuming and costly biochemical experiments, and it is difficult to predict toxicity. Therefore, it is of great significance to develop a fast, accurate and low-cost method to evaluate and predict toxicity changes during its degradation process.

[0003] Machine learning technology has broad application potential and significant advantages in the field of environmental pollution treatment. Its powerful data collection and analysis capabilities enable it to quickly and accurately identify known or unknown pollutants. In addition, machine learning technology can also consider the impact of multiple conditions on environmental reactions, establish pollutant removal kinetics and reaction condition models, and thus improve the efficiency and resource utilization of pollutant removal processes. By integrating large-scale observational data and machine learning technology, the accuracy and speed of data analysis can be improved, providing a more effective method for environmental management and governance.

[0004] Toxicology testing and quantitative structure-activity relationship simulation are two common methods for evaluating the toxicity of chemical substances. Toxicology testing is carried out at various levels such as individual and acute toxicity. Although intuitive toxicity data can be provided, testing often requires a lot of resources and time, and because it involves biological experiments, it may be restricted in law and ethics. A large number of studies have been conducted on quantitative structure-activity relationship simulation, including the use of TEST and ECOSAR software for toxicity simulation, which can be analyzed and calculated by inputting the chemical structure of the substance. It has been widely used in toxicity simulation in the study of tetracycline intermediates in photocatalytic degradation. However, the degradation process is relatively complex, and the types and quantity changes of intermediates are difficult to summarize. Therefore, the research lacks a comprehensive toxicity simulation prediction of the solution of the photocatalytic reaction base solution. Summary of the invention

[0005] In order to solve the problems that toxicological biological testing is costly and time-consuming and there is a lack of a model for the changes in the types and quantities of intermediate products in the photocatalytic degradation process, the purpose of the present invention is to provide a toxicity prediction method for the photocatalytic degradation process of tetracycline based on machine learning, which can significantly reduce toxicological test verification experiments, is simple and easy to operate, saves resources and processing time, and improves processing efficiency.

[0006] To achieve the above object, the present invention adopts the following technical solution: a method for predicting the toxicity of tetracycline photocatalytic degradation process based on machine learning, the method comprising the following steps in order:

[0007] (1) Identify the more important intermediates in the photocatalytic degradation of tetracycline;

[0008] (2) inferring the reaction pathway of the photocatalytic degradation of tetracycline based on the intermediate product determined in step (1);

[0009] (3) The acute toxicity of the simulated intermediates to aquatic organisms was used as the acute toxicity evaluation standard and quantified using the median lethal concentration;

[0010] (4) Determine the types and contents of intermediates under different conditions such as initial tetracycline concentration, catalyst dosage, and light reaction time, and establish a machine learning model for the types and contents of intermediates under different conditions;

[0011] (5) determining the initial reaction conditions, wherein the initial reaction conditions include initial concentration, catalyst dosage, and light irradiation reaction time, inputting the data to be predicted into the machine learning model, predicting the intermediate products and contents during the photocatalytic degradation of tetracycline, and obtaining predicted contents of different intermediate products;

[0012] (6) The toxicity of tetracycline during photocatalytic degradation is predicted by using the acute toxicity evaluation standard obtained in step (3) and the predicted contents of different intermediate products obtained in step (5).

[0013] In step (1), the more important intermediate products in the process of photocatalytic degradation of tetracycline are determined by high performance liquid chromatography-mass spectrometry, the ion fragment scanning range of the high performance liquid chromatography-mass spectrometry is 100m / z to 500m / z, and 8 to 12 intermediate products are selected.

[0014] In step (3), the 48-hour half-lethal concentration of simulated Daphnia magna is selected as the acute toxicity evaluation standard.

[0015] In step (4), the initial concentration ranges from 10 ppm to 50 ppm.

[0016] In step (4), the catalyst dosage ranges from 0.2 g / L to 1.0 g / L.

[0017] In step (4), the illumination reaction time ranges from 5 min to 30 min.

[0018] In step (4), the establishment of a machine learning model for the type and content of intermediate products under different conditions specifically refers to: the machine learning model adopts a random forest machine learning model, and collects tetracycline concentrations of 10 mg / L, 20 mg / L, 30 mg / L, 40 mg / L, and 50 mg / L by high performance liquid chromatography-mass spectrometry; catalyst dosages of 0.2 g / L, 0.4 g / L, 0.6 g / L, 0.8 g / L, and 1.0 g / L; the change in the quantity of the selected intermediate product is tested at light reaction times of 5 min, 10 min, 15 min, 20 min, 25 min, and 30 min; in the obtained data, the type of intermediate product, TiO2 concentration, TC initial concentration, and light reaction time are used as model input variables, and the peak area is used as the model output variable to establish a random forest machine learning model.

[0019] It can be seen from the above technical scheme that the beneficial effects of the present invention are: first, the present invention can significantly reduce toxicological test verification experiments in toxicity simulation assessment, is simple and easy, conforms to the concept of green environmental protection, and can provide important support for the tetracycline degradation process; second, the present invention is based on a machine learning model, which can simulate and predict the content of intermediate products under different reaction conditions, and combine it with toxicity simulation for evaluation, which helps experimenters to select reaction stages with lower toxicity under different conditions, save resources and processing time, and improve processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the method of the present invention;

[0021] Figure 2 is the total ion current chromatogram of the reaction substrate;

[0022] Figure 3 It is the reaction path diagram of the degradation process;

[0023] Figure 4 Simulated acute toxicity graphs for tetracycline and intermediates;

[0024] Figure 5 A scatter plot of the regression results of the actual value and the predicted value of the established model;

[0025] Figure 6 The simulated comprehensive toxicity diagram of solutions with different reaction times. DETAILED DESCRIPTION

[0026] like Figure 1 As shown, a method for predicting the toxicity of tetracycline photocatalytic degradation process based on machine learning includes the following steps in order:

[0027] (1) Identify the more important intermediates in the photocatalytic degradation of tetracycline;

[0028] (2) inferring the reaction pathway of the photocatalytic degradation of tetracycline based on the intermediate product determined in step (1);

[0029] (3) Use TEST software to simulate the acute toxicity of the intermediates to aquatic organisms as the acute toxicity evaluation standard and use the median lethal concentration for quantification;

[0030] (4) Determine the types and contents of intermediates under different conditions such as initial tetracycline concentration, catalyst dosage, and light reaction time, and establish a machine learning model for the types and contents of intermediates under different conditions;

[0031] (5) determining the initial reaction conditions, wherein the initial reaction conditions include initial concentration, catalyst dosage, and light irradiation reaction time, inputting the data to be predicted into the machine learning model, predicting the intermediate products and contents during the photocatalytic degradation of tetracycline, and obtaining predicted contents of different intermediate products;

[0032] (6) The toxicity of tetracycline during photocatalytic degradation is predicted by using the acute toxicity evaluation standard obtained in step (3) and the predicted contents of different intermediate products obtained in step (5).

[0033] In step (1), the more important intermediate products in the process of photocatalytic degradation of tetracycline are determined by high performance liquid chromatography-mass spectrometry, i.e., LCMS. The ion fragment scanning range of high performance liquid chromatography-mass spectrometry is 100m / z to 500m / z, and 8 to 12 intermediate products are selected. The initial concentration of tetracycline is 20ppm, the catalyst dosage is 30mg, and the illumination time is 5min as the photocatalytic reaction conditions. The intermediate products are tested and identified using the MS1 SCAN mode in high performance liquid chromatography-mass spectrometry, such as Figure 2 As shown, 10 relatively obvious substance peaks were detected, among which m / z=445.25 was the characteristic peak of tetracycline, and the remaining 9 peaks were substance peaks corresponding to the intermediate products.

[0034] The step (2) specifically refers to: Figure 3 As shown, the photocatalytic degradation pathway of TC catalyzed by P25 under full spectrum irradiation: the generation of P1 (m / z=459) in the left pathway involves multiple hydroxylation reactions and dehydration reactions on the C ring, and then undergoes deamidation and dihydroxylation to generate P2 (m / z=340); then, the A ring is opened and hydrogenated to form P3 (m / z=246), which is further decomposed to obtain P4 (m / z=150);

[0035] In the right pathway, under the action of various active oxidizing substances, a deamidation reaction occurs on the A ring. Due to the low binding energy of NC, the NC bond on the A ring is broken. In addition, a series of demethylation, double bond breaking, and ring-opening reactions occur on the A ring to form P5 (m / z=318). After a Dakin-like reaction, the acetyl group on the B ring is replaced by a hydroxyl group to form P6 (m / z=274). After the B ring in P6 is opened, it is demethylated to form P7 (m / z=218); then it is demethylated to form P8 (m / z=191). After the BCD rings in P6 are destroyed, P9 (m / z=128) is formed.

[0036] In step (3), the 48-hour half-lethal concentration of simulated Daphnia magna is selected as the acute toxicity evaluation standard.

[0037] In step (4), the initial concentration ranges from 10 ppm to 50 ppm.

[0038] In step (4), the catalyst dosage ranges from 0.2 g / L to 1.0 g / L.

[0039] In step (4), the illumination reaction time ranges from 5 min to 30 min.

[0040] In step (4), the establishment of a machine learning model for the type and content of intermediate products under different conditions specifically refers to: the machine learning model adopts a random forest machine learning model, and collects tetracycline concentrations of 10 mg / L, 20 mg / L, 30 mg / L, 40 mg / L, and 50 mg / L by high performance liquid chromatography-mass spectrometry; 0.2 g / L, 0.4 g / L, 0.6 g / L, 0.8 g / L, and 1.0 g / L of catalyst dosage; the change in the quantity of the selected intermediate product is tested every 5 min, 10 min, 15 min, 20 min, 25 min, and 30 min of light reaction time; in the obtained data, the type of intermediate product, TiO2 concentration, TC initial concentration, and light reaction time are used as model input variables, and the peak area is used as the model output variable to establish a random forest machine learning model. In order to improve the efficiency of training, the data is normalized, and 80% is set as a training set and 20% is set as a test set. The grid search method is used to screen and optimize the hyperparameters, and the cross-validation is used to score the hyperparameter combination. The optimized hyperparameters are as follows: n_estimators = 100, max_depth = 50, max_features = 20, random_state = 2, which are used in the subsequent random forest machine learning model calculation process.

[0041] Embodiment 1

[0042] S1: Add 30 mg of P25 to 50 ml of 20 mg / L TC solution, stir in the dark for 30 minutes, and then place under a 300W xenon lamp (280nm<λ<980nm) for photocatalytic reaction. Take a sample of the solution from the reactor, filter it through a 0.2 μm filter, and inject it into the LCMS. Select the intermediate with a high content and obvious characteristic peaks in the LCMS total ion flow chart, such as Figure 2 As shown, the primary and secondary mass spectrometry scanning was performed, and the results were searched and compared with the mass spectrometry database MassBank and other databases to determine the more important intermediates in the photocatalytic degradation of tetracycline;

[0043] S2: Based on the changes in the content of the intermediate products determined in step S1 and taking into account the influence of active oxidizing substances, the reaction pathway of the photocatalytic degradation of tetracycline is proposed, such as Figure 3 As shown;

[0044] S3: Output the chemical structure of the intermediate product identified in step S1 using the chemical structure drawing tool software ChemDraw and simulate and quantify the acute toxicity of the 48-hour half-lethal concentration of Daphnia magna using the TEST software, such as Figure 4 As shown;

[0045] S4: The influence of reaction conditions such as 10mg / L, 20mg / L, 30mg / L, 40mg / L, 50mg / L of tetracycline concentrations, 0.2g / L, 0.4g / L, 0.6g / L, 0.8g / L, 1.0g / L of catalyst dosages, and 5min, 10min, 15min, 20min, 25min, 30min of light reaction time on the quantity changes of tetracycline and nine intermediates selected in step S1 was collected by LCMS determination, and a random forest machine learning model was established, such as Figure 5 As shown, the correlation coefficient of the random forest machine learning model is 0.94, and the random forest machine learning model is well established;

[0046] S5: Testing a solution with an unknown tetracycline concentration to obtain its initial concentration, catalyst dosage, and light irradiation reaction time for test control, and inputting them into the random forest machine learning model established in step S4 to obtain the type and quantity of the intermediate product determined in step S1;

[0047] S6: Take the negative logarithm of the toxicity prediction value simulated in step S3 with the base 10, multiply it by the corresponding content of the intermediate product obtained in step S5, and then add them together to obtain the simulated prediction value of the comprehensive toxicity of the reaction base liquid during the degradation process, such as Figure 6 shown.

[0048] In summary, the present invention can significantly reduce toxicological test verification experiments in toxicity simulation assessment, is simple and easy, conforms to the concept of green environmental protection, and can provide important support for the tetracycline degradation process; the present invention is based on a machine learning model, which can simulate and predict the content of intermediate products under different reaction conditions, and combine it with toxicity simulation for evaluation, which helps experimenters to select reaction stages with lower toxicity under different conditions, save resources and processing time, and improve processing efficiency.

[0049] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning, characterized in that: The method comprises the following steps in the following order: (1) Identify the more important intermediates in the photocatalytic degradation of tetracycline; (2) inferring the reaction pathway of the photocatalytic degradation of tetracycline based on the intermediate product determined in step (1); (3) The acute toxicity of the simulated intermediates to aquatic organisms was used as the acute toxicity evaluation standard and quantified using the median lethal concentration; (4) Determine the types and contents of intermediates under different conditions such as initial tetracycline concentration, catalyst dosage, and light reaction time, and establish a machine learning model for the types and contents of intermediates under different conditions; (5) determining the initial reaction conditions, wherein the initial reaction conditions include initial concentration, catalyst dosage, and light irradiation reaction time, inputting the data to be predicted into the machine learning model, predicting the intermediate products and contents during the photocatalytic degradation of tetracycline, and obtaining predicted contents of different intermediate products; (6) The toxicity of tetracycline during photocatalytic degradation is predicted by using the acute toxicity evaluation standard obtained in step (3) and the predicted contents of different intermediate products obtained in step (5).

2. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (1), the more important intermediate products in the process of photocatalytic degradation of tetracycline are determined by high performance liquid chromatography-mass spectrometry, the ion fragment scanning range of the high performance liquid chromatography-mass spectrometry is 100m / z to 500m / z, and 8 to 12 intermediate products are selected.

3. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (3), the 48-hour half-lethal concentration of simulated Daphnia magna is selected as the acute toxicity evaluation standard.

4. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (4), the initial concentration ranges from 10 ppm to 50 ppm.

5. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (4), the catalyst dosage ranges from 0.2 g / L to 1.0 g / L.

6. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (4), the illumination reaction time ranges from 5 min to 30 min.

7. The method for predicting toxicity of tetracycline photocatalytic degradation process based on machine learning according to claim 1, characterized in that: In step (4), the establishment of a machine learning model for the type and content of intermediate products under different conditions specifically refers to: the machine learning model adopts a random forest machine learning model, and collects tetracycline concentrations of 10 mg / L, 20 mg / L, 30 mg / L, 40 mg / L, and 50 mg / L by high performance liquid chromatography-mass spectrometry; catalyst dosages of 0.2 g / L, 0.4 g / L, 0.6 g / L, 0.8 g / L, and 1.0 g / L; the change in the quantity of the selected intermediate product is tested at light reaction times of 5 min, 10 min, 15 min, 20 min, 25 min, and 30 min; in the obtained data, the type of intermediate product, TiO2 concentration, TC initial concentration, and light reaction time are used as model input variables, and the peak area is used as the model output variable to establish a random forest machine learning model.

Images