QSAR (Quantitative Synthetic Aperture Radar) model for predicting adsorption of microplastics to organic pollutants, construction method and application

Abstract

The invention belongs to the technical field of environmental science and chemical engineering, and provides a QSAR (Quantitative Synthetic Aperture Radar) model for predicting adsorption of microplastics to organic pollutants, a construction method and application. A quantitative structure-activity relationship is constructed based on drug physicochemical properties and linear free energy parameters, the ability of polyethylene and polystyrene micro-plastics to adsorb organic pollutants, especially non-steroidal anti-inflammatory drugs, before and after aging is predicted, and the ability of the polyethylene and polystyrene micro-plastics to adsorb organic pollutants, especially non-steroidal anti-inflammatory drugs, is predicted by taking a distribution coefficient (Kd) of the organic pollutants in the micro-plastics as a dependent variable and taking the physicochemical property parameters of the organic pollutants as independent variables. A retreating stepwise regression method is adopted to screen variables and determine model parameters, on the premise of ensuring model interpretability, an optimal independent variable combination is screened out, the internal relation between the variables is accurately recognized, redundancy with high correlation is effectively eliminated, the collinearity degree between the variables is remarkably reduced, and the method is suitable for large-scale popularization and application. The obtained multiple QSAR model formulas are used for predicting organic pollutants adsorbed by micro-plastics, especially non-steroidal anti-inflammatory drugs, and the prediction accuracy and precision are improved.

Description

Technical Field

[0001] This invention relates to the fields of environmental science and chemical engineering technology, specifically to a QSAR model, construction method, and application for predicting the adsorption of organic pollutants, especially nonsteroidal anti-inflammatory drugs, by microplastics. Background Technology

[0002] Plastics are high-molecular polymers synthesized from resins. Due to their low cost, ease of manufacturing, high chemical stability, and good electrical insulation, they are widely used in social production and daily life. It is estimated that by 2050, 12 billion tons of plastic waste will be landfilled and released into the natural environment. It has been reported that more than 250,000 tons of microplastics are released into marine systems annually, and microplastics have also been detected in lakes, rivers, and drinking water. Under natural conditions, waste plastics undergo a complex physical, chemical, and biological aging process under the influence of various environmental factors (such as ultraviolet light, wind, heat, pH, and microorganisms), gradually breaking down into microplastics and even nanoplastics. These fragments exhibit significant environmental mobility, accumulating, migrating, and spreading in different ecosystems. Furthermore, the aging process significantly affects the adsorption mechanism of microplastics for organic pollutants and their enrichment capacity, significantly increasing the complexity of the adsorption behavior of microplastics and pollutants. Therefore, systematically studying the aging mechanism of microplastics and their adsorption mechanism for pollutants is of great scientific significance for accurately assessing the environmental behavior and ecological risks of microplastics.

[0003] Among the five most produced and consumed plastics in the world today are polyethylene (PE) and polystyrene (PS). Both PE and PS are high molecular weight compounds formed by the polymerization of numerous monomers. Their molecular chains have high molecular weights and are composed of many repeating structural units. The monomer of PE microplastics is ethylene, with the structural formula CH2=CH2. Its molecular chain is relatively regular and simple, exhibiting a linear structure. Carbon atoms in the chain are linked by single bonds, and hydrogen atoms are evenly distributed around the carbon atoms. This structure makes the PE molecular chain chemically stable, difficult to directly chemically modify, and its intermolecular forces are relatively weak. The monomer of PS microplastics is styrene, with the structural formula C6H5-CH=CH2. Due to the presence of the benzene ring, the regularity of the molecular chain is somewhat affected. The benzene ring is a large conjugated system with high rigidity, giving polystyrene some unique chemical properties and stronger intermolecular forces. This results in PS microplastics having higher hardness, brittleness, and rigidity, and their mechanical and chemical properties are significantly different from those of PE microplastics.

[0004] Non-steroidal anti-inflammatory drugs (NSAIDs) possess multiple pharmacological effects, including antipyretic, analgesic, anti-inflammatory, and anticoagulant properties. Approximately 35 million people rely on these drugs daily to relieve inflammation and pain. Several NSAIDs, such as diclofenac, ibuprofen, and naproxen, have been reported to have significant negative physiological effects on aquatic organisms, specifically increased mortality, endocrine system disorders, abnormal growth and development, and reproductive dysfunction. Notably, in aquatic environments with pH 7, NSAIDs primarily exist in anionic form, which is difficult to remove from water using conventional physical methods (such as coagulation, sedimentation, and filtration). Microplastics and NSAIDs, as two emerging pollutants widely coexisting in the environment, have potential impacts on environmental quality, ecosystems, and the health of animals and humans. However, current research on their interactions and environmental impacts is relatively limited. Furthermore, existing model studies mainly focus on a limited range of organic pollutants, such as polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), and antibiotics, lacking broader data support. Meanwhile, there are few, or even no, studies on QSAR models for the adsorption of nonsteroidal anti-inflammatory drugs (NSAIDs) by microplastics. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a QSAR model construction method for predicting the adsorption of organic pollutants by microplastics. Based on the physicochemical properties of drugs and linear free energy parameters, a quantitative structure-activity relationship is constructed to predict the ability of polyethylene and polystyrene microplastics to adsorb non-steroidal anti-inflammatory drug-type organic pollutants before and after aging. The method includes the following steps:

[0006] 1) Obtain the adsorption coefficient K of organic pollutants in microplastics. d value;

[0007] 2) Obtain chemical descriptors for organic pollutants, including: octanol-water partition coefficient (LogKow), dissociation constant (pKa), excess molecular molar refractive index (E), molecular dipole / polarizability (S), hydrogen bond acidity (A), hydrogen bond basicity (B), molecular volume (V), and electrostatic Coulomb force parameter (J). - );

[0008] 3) Take the K obtained in step 1) d Using the values ​​as the dependent variable and the chemical descriptors obtained in step 2) as independent variables, a QSAR model is constructed using stepwise multiple linear regression; the general formula for the QSAR model predicting the adsorption of organic pollutants by microplastics is obtained:

[0009] LogK d=N1Z1+N2Z2+N3Z3+N4Z4+N5Z5+N6Z6+N7Z7+N8Z8+C; where Z i N is one of the eight descriptor arguments mentioned in step 2); i C is the coefficient; C is the constant term.

[0010] The optimal QSAR model is obtained by automatically backtracking and removing independent variables using SPSS. Finally, the optimal prediction model obtained in step 3) is validated using leave-one-out cross-validation. 2 LOO To test the robustness of the QSAR model obtained in step 3), Williams plots are used to assess the applicability of the QSAR model obtained in step 3).

[0011] Furthermore, step 1) specifically involves:

[0012] 1.1 The equilibrium adsorption capacity q was obtained through isothermal adsorption experiments. e ,

[0013] 1.2 Obtain K through further calculation d Value, K d The specific formula for calculating the value is as follows:

[0014]

[0015] Among them, K d The unit is L / Kg;

[0016] q e The equilibrium adsorption capacity (μmol / Kg) of organic pollutants in the microplastic phase;

[0017] C0 represents the initial concentration (μmol / L) of the organic pollutant;

[0018] Ce represents the equilibrium concentration (μmol / L) of organic pollutants in the aqueous phase.

[0019] V is the volume (L) of the aqueous solution in the adsorption experiment.

[0020] Further, step 3) specifically involves: first, performing Pearson correlation analysis on the chemical descriptors obtained in step 2) using SPSS to screen out variables with significant linear overlap; then, using the K obtained in step 1) dThe values ​​and chemical descriptors obtained in step 2) are used as dependent and independent variables, respectively. The model is built using the stepwise regression method in the multiple linear regression (MLR) algorithm in SPSS (23.0) software. Predictive variables are introduced or removed stepwise based on statistical significance criteria. In each iteration, the algorithm selects the optimal combination of variables that minimizes the sum of squared residuals. By stepwise optimization, the explanatory power and prediction accuracy of the model are improved, and the QSAR model for predicting the adsorption of organic pollutants by microplastics is obtained.

[0021] Secondly, this invention provides a QSAR model for predicting the adsorption of organic pollutants by microplastics. Based on the above construction method, the formula for the QSAR model is as follows:

[0022] LogK d =N1Z1+N2Z2+N3Z3+N4Z4+N5Z5+N6Z6+N7Z7+N8Z8+C

[0023] Z i N is one of the eight descriptor arguments mentioned in step 2); i C is the coefficient; C is the constant term.

[0024] This invention provides the above-mentioned construction method and QSAR model for the application of microplastic adsorption prediction of organic pollutants.

[0025] Furthermore, the QSAR model for predicting the adsorption of organic pollutants by microplastics also includes: when the organic pollutant is a nonsteroidal anti-inflammatory drug, and,

[0026] When the microplastic is the original PE microplastic, the QSAR model formula (1) is:

[0027] LogK d =0.450B-1.039J - +1.828(1)

[0028] When the microplastic is a PE-UV microplastic, the QSAR model formula (2) is:

[0029] LogK d =0.431B - 0.990J - +1.922(2)

[0030] When the microplastic is PE-HT microplastic, the QSAR model formula (3) is:

[0031] LogK d =0.344B - 0.875J - +1.859(3)

[0032] When the microplastic is the original PS microplastic, the QSAR model - formula (4) is:

[0033] LogK d =0.415B - 0.837J - -0.105pKa + 1.939(4)

[0034] When the microplastic is a PS-UV microplastic, the QSAR model is given by formula (5):

[0035] LogK d =0.372B-0.556V-1.579A+0.863(5)

[0036] When the microplastic is PS-HT microplastic, the QSAR model is given by formula (6):

[0037] LogK d =0.508B-0.697V-1.786A+0.566(6)

[0038] Finally, this invention also provides the application of the above six QSAR models for predicting the adsorption of nonsteroidal anti-inflammatory drugs by microplastics.

[0039] This invention uses the microplastic adsorption coefficient K d Using the physicochemical properties of organic pollutants, especially nonsteroidal anti-inflammatory drugs (NSAIDs), as dependent variables and independent variables, a backward stepwise regression method was used to screen variables and determine model parameters. The chemical structural formulas and names of ten NSAIDs were obtained from EPI suite (4.1), and the chemical structures of each drug were substituted into the KOWWIN module to calculate the octanol-water partition coefficient (LogKow) and dissociation constant (pKa) of each drug. LogKow represents the relative ratio of the drug's lipid solubility to its water solubility, mainly reflecting the drug's hydrophobicity; pKa reflects the degree of dissociation of the drug under a specific pH environment.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] (1) This invention constructs a structure-activity relationship (SPR) for the partitioning of organic pollutants, especially nonsteroidal anti-inflammatory drugs (NSAIDs), on the surface of microplastics in an aquatic environment, considering and simulating the adsorption of pollutants by microplastics after aging. The construction method provided in this invention targets PE and PS microplastics as representative examples, and the final model is obtained by evaluating and adjusting the two aging methods separately. Existing technologies usually only consider a single aging method because they ignore the influence of different aging methods on the model. The construction of the model in this application includes the step of evaluation and adjustment, and also illustrates the influence of different aging methods on the model; for example, the aging method of PE affects the coefficient of the variable, while the aging method of PS affects the selection of the variable. The constructed SPR starts from the reaction mechanism and expands the application scope of the QSAR model for organic pollutants. It can not only further expand the application domain of the research objects and models, but also analyze the effect of the physicochemical properties of NSAIDs on K d The influence mechanism of the value and even the prediction of K values ​​of NSAIDs on microplastics d This value is significant for a more comprehensive understanding of the migration and transformation of microplastic-NSAID complexes under aquatic conditions.

[0043] (2) The construction method provided in this application has a wide range of variable selection, covering various molecular weights and high coverage, making the method scalable. Furthermore, the selection of different independent variables for specific microplastics also provides specificity. The parameters selected in this invention are more representative, based on the partitioning behavior of organic matter in different phases and possible mechanisms of action. Representative parameters are selected at the drug molecule level, such as hydrogen bond acidity (A), hydrogen bond basicity (B), and electrostatic coulombic force (J). - The prediction equations were optimized, making the prediction model equations simple and reliable, which is more conducive to evaluating the distribution behavior and mechanism of action of nonsteroidal anti-inflammatory drugs in microplastics.

[0044] (3) The regression analysis method of this invention is simple and can quickly predict the combined pollution of microplastics and nonsteroidal anti-inflammatory drugs (NSAIDs) with limited experimental data. This is beneficial for screening out NSAID-related organic pollutants that need to be prioritized for control in the environment. Under the premise of ensuring model interpretability, it can screen out the optimal combination of independent variables, thereby more accurately identifying the intrinsic relationship between independent variables and response variables; it can effectively eliminate redundant variables that are highly correlated with other independent variables, thereby significantly reducing the degree of collinearity among variables and improving the prediction accuracy of the model. Attached Figure Description

[0045] Figure 1 LogK is the adsorption model of nonsteroidal anti-inflammatory drugs for six microplastics in this invention. d A graph showing the fit between predicted and experimental values;

[0046] Figure 2The Williams application domain diagram for the six microplastic adsorption models of nonsteroidal anti-inflammatory drugs of the present invention is used to determine whether there are outliers in the model and whether the predicted values ​​are within the valid range. Detailed Implementation

[0047] Example 1: Taking ten nonsteroidal anti-inflammatory drugs as examples, the QSAR model construction method for predicting the adsorption of organic pollutants by microplastics is as follows:

[0048] 1) Microplastic adsorption coefficient K of nonsteroidal anti-inflammatory drugs d Obtaining the equilibrium adsorption capacity q: The equilibrium adsorption capacity q was obtained through isothermal adsorption experiments. e Further calculations yielded K. d value;

[0049] 40 mg of MPs was weighed into 50 mL centrifuge tubes, followed by the addition of 40 mL of nonsteroidal anti-inflammatory drug (NSAID) solutions with concentrations of 0.005, 0.01, 0.02, 0.04, 0.05, 0.08, and 0.1 mmol / L, respectively. The centrifuge tubes were shaken at 150 rpm for 24 h at 25 °C in the dark. The supernatant from the centrifuge tubes was aspirated using a 1 mL syringe and filtered through a 0.45 μm microporous membrane for solid-liquid separation. The peak area of ​​the NSAID was detected by high-performance liquid chromatography (HPLC), and the equilibrium concentration of the NSAID was calculated. The equilibrium adsorption capacity (μmol / Kg) and initial concentration (μmol / L) of the NSAID in the microplastic phase and in the aqueous phase were calculated. Two parallel samples were used for each experiment, and the average value was taken.

[0050] To investigate the partitioning characteristics of nonsteroidal anti-inflammatory drugs (NSAIDs) as organic pollutants in a microplastic-freshwater system, the equilibrium partition coefficients (K0) of 10 NSAIDs on the surfaces of PE and PS microplastics were experimentally determined. d value)

[0051]

[0052] In the formula, K d The unit is L / Kg;

[0053] q e The equilibrium adsorption capacity (μmol / Kg) of nonsteroidal anti-inflammatory drugs in the microplastic phase;

[0054] C0 is the initial concentration (μmol / L) of the nonsteroidal anti-inflammatory drug;

[0055] C e This represents the equilibrium concentration (μmol / L) of the nonsteroidal anti-inflammatory drug in the aqueous phase.

[0056] V is the volume (L) of the aqueous solution in the adsorption experiment.

[0057] LogK experiment of nonsteroidal anti-inflammatory drugs d surface

[0058]

[0059] LogK in this table d These are experimentally detected values, used for model building later.

[0060] 2) Collection of chemical descriptors; the chemical structural formulas and chemical names of ten nonsteroidal anti-inflammatory drugs (NSAIDs) were obtained from EPI suite (4.1), and the chemical structures of each drug were substituted into its KOWWIN module to calculate the octanol-water partition coefficient LogKow, which represents the relative ratio of the drug's lipid solubility to water solubility, mainly reflecting the drug's hydrophobicity; pKa reflects the degree of dissociation of the drug under a specific pH environment. This includes the octanol-water partition coefficient (LogKow), dissociation constant (pKa) of ten NSAIDs, and various linear free energy parameters obtained from the UFZ-LSER database and literature: excess molecular molar refractive index (E), molecular dipole / polarizability (S), hydrogen bond acidity (A), hydrogen bond basicity (B), and molecular volume (V); in particular, the electrostatic coulomb force parameter (J) was introduced. - This characterizes the electrostatic interaction between nonsteroidal anti-inflammatory drugs and microplastics.

[0061] The multi-parameter linear free energy equation and related parameter system proposed by Abraham in 2004 are adopted, including: 1. Excess molecular molar refractive index (E), characterizing the n- / π- electron pair interaction and dispersion force between adsorbate and adsorbent; 2. Molecular dipole / polarizability (S), reflecting the polar / polarization interaction between molecules; 3. Hydrogen bond acidity (A) and hydrogen bond basicity (B), describing the characteristics of hydrogen bond interaction; 4. Molecular volume (V), characterizing the steric hindrance effect; In particular, the electrostatic Coulomb force parameter (J) is introduced. - This reflects and describes the electrostatic interactions between anions in a specific system;

[0062] Descriptor parameter table

[0063]

[0064] 3) Predictive model establishment; the specific method is as follows: first, SPSS is used to perform Pearson correlation analysis on each chemical descriptor to screen out variables with significant linear overlap, and then LogK d The values, physicochemical properties, and linear free energy parameters were used as dependent and independent variables, respectively. Then, stepwise multiple linear regression (using the backward regression method) was used to construct the QSAR model, thereby obtaining a series of quantitative structure-activity relationships.

[0065] The general formula for the QSAR model to predict the adsorption of nonsteroidal anti-inflammatory drugs by microplastics is obtained as follows:

[0066] LogK d =N1Z1+N2Z2+N3Z3+N4Z4+N5Z5+N6Z6+N7Z7+N8Z8+C

[0067] Z i N is one of the eight descriptor arguments mentioned in step 2); i C is the coefficient; C is the constant term.

[0068] The goodness of fit of the QSAR model established above is determined by the coefficient of determination (R²). 2 Adjusted coefficient of determination (R) 2 adj The root mean square error (RMSE) is used to test for independence. The regression model is evaluated using the p-value in analysis of variance (ANOVA). A p-value less than 0.05 indicates that at least one independent variable in the equation can be used to predict the dependent variable below the 95% significance level. The variance inflation factor (VIF) is used to test for multicollinearity or correlation among the independent variables. A higher VIF value indicates a higher correlation with one or more of the remaining independent variables. Generally, when the VIF value is greater than 10, the independent variables are considered correlated. Therefore, the VIF of the descriptor should be <10, and the tolerance should be greater than 0.1 to avoid multicollinearity issues. The Durbin-Watson value (DW value) is used to test the independence of the residuals in the multiple linear regression model. The DW value is an important statistical indicator used in statistical analysis to assess the first-order autocorrelation of residuals. Its value generally ranges from 0 to 4. When the Durbin-Watson value is within this range, the data are generally considered independent.

[0069] The applicability of the developed model is verified using Williams plots, which depict the relationship between standardized residuals and leverage values. Outliers (abnormal Y values) are identified through standardized residuals, typically those greater than three standard deviations. The leverage point (abnormal X values) is determined by the leverage value; if the leverage value > h*, the warning value is h* = 3(p+1) / N, where p is the number of model parameters and N is the sample size. Leave-one-out cross-validation (Q-test) is used. 2 LOO ) Test the robustness of the model fit, Q 2 LOO A score >0.5 indicates that the constructed model is good. Q 2 LOO The principle is to exclude data points one by one, use the remaining data points for model development, and then check the predictability of the excluded data points by comparing the calculated values.

[0070] Example 2: The original PE QSAR model established according to the method in Example 1

[0071] In Model 1, R 2 =0.975, predictors are pKa, LogKow, E, S, B, A, V, J - However, none of the variables passed the significance test (all greater than 0.05), and except for pKa and LogKow, E, S, B, A, V, and J... - None of the variables passed the multicollinearity diagnostic test, indicating that there is a multicollinearity problem among the model parameters (VIF>10);

[0072] Furthermore, in Model 2, variables A and R are removed by going back. 2 =0.971, predictor variable is J - LogKow, pKa, E, B, V, S, but none of the variables passed the significance test (all greater than 0.05), and except for pKa and LogKow, E, S, B, V, J - None of the variables passed the collinearity diagnostic test (VIF>10);

[0073] Furthermore, in Model 3, the variable pKa and R are removed backwards. 2 =0.966, predictor variable is J - LogKow, E, B, V, S, but none of the variables passed the significance test (all greater than 0.05), and except for LogKow, E, S, B, V, J - None of the variables passed the collinearity diagnostic test (VIF>10);

[0074] Furthermore, in Model 4, the variable LogKow is removed after backwards. 2 =0.911, predictor variable is J - E, B, V, S, but none of the variables passed the significance test (all greater than 0.05) and the collinearity diagnostic test (VIF>10);

[0075] Furthermore, in Model 5, variables V and R are removed by going back. 2 =0.883, predictor variable is J - E, B, S, variable J - Variable B passed the significance test (both were 0.003 < 0.05), while variables E and S did not pass the significance test (both were greater than 0.05), and all variables failed the collinearity diagnostic test (VIF > 10).

[0076] Furthermore, in Model 6, variables S and R are removed by going back. 2 =0.862, predictor variable is J - E, B, variable J- Variable B passed the significance test (0.001 and 0.002 respectively), while variable E did not pass the significance test. All three variables passed the collinearity diagnostic test (VIF<10).

[0077] Furthermore, in Model 7, variables E and R are removed by going back. 2 =0.856, predictor variable is J - B, all variables passed the significance test (all less than 0.05) and the collinearity diagnostic test (VIF = 8.473 < 10);

[0078] Summary of Stepwise Optimization of PE QSAR Model

[0079]

[0080]

[0081] a. Predictor variables: (constant), J - ,A,LogKow,pKa,E,B,V,S

[0082] b. Predictor variables: (constant), J - ,LogKow,pKa,E,B,V,S

[0083] c. Predictor variables: (constant), J - ,LogKow,E,B,V,S

[0084] d. Predictor variables: (constant), J - ,E,B,V,S

[0085] e. Predictor variables: (constant), J - E,B,S

[0086] f. Predictor variables: (constant), J - E,B

[0087] g. Predictor variables: (constant), J - B

[0088] h. Dependent variable: PE-LogK d

[0089]

[0090]

[0091] Reaction mechanism analysis

[0092] Parameter table of the established optimal model

[0093]

[0094] Parameter table of the established optimal model

[0095]

[0096] Note: Dependent variable = PE - LogK d

[0097] *p<0.05, **p<0.01

[0098] Eight descriptors were selected (pKa, LogKow, E, S, B, A, V, J). - ) as independent variables, and compare them with PE-LogK d Stepwise multiple linear regression analysis was performed. After automatic model identification and stepwise optimization, the final remaining term descriptors (B, J) were obtained. - ).

[0099] The optimal QSAR model constructed is:

[0100] LogK d =0.450B-1.039J - +1.828

[0101] J in the model - Having the largest absolute value of t (t = -6.435) indicates that J - This is the main factor affecting the adsorption capacity of virgin PE microplastics for nonsteroidal anti-inflammatory drugs. (Descriptor J) - This reflects the electrostatic Coulombic force of the compound, indicating that the adsorption of nonsteroidal anti-inflammatory drugs (NSAIDs) by the nonpolar PE is influenced by electrostatic interactions. Descriptor J - The regression coefficient was -1.039 (p = 0.001 < 0.01), indicating that J... - Will be PE-LogK d A significant negative relationship was observed. This indicates that the smaller the anion coulombic interaction force of nonsteroidal anti-inflammatory drugs (NSAIDs), the lower the LogK... d The higher the value, the stronger the adsorption capacity of PE microplastics for nonsteroidal anti-inflammatory drugs (NSAIDs). This is because the negatively charged surface of PE microplastics generates electrostatic repulsion with the negative charges between NSAID molecules. Drugs with weaker anionic Coulombic interactions among the ten NSAID molecules exhibit even weaker electrostatic repulsion when adsorbed with PE, thus displaying a higher LogK value. dThe value. Descriptor B is the second most important descriptor in the model (t = 5.811). Descriptor B represents the drug's ability to act as a hydrogen bond acceptor; the larger the value of B, the stronger the drug's ability to bind to hydrogen bond donor groups on the microplastic surface. The regression coefficient of B in the formula is 0.450 (p = 0.001 < 0.01), indicating that B has an impact on the PE-LogK... d A significant positive correlation was observed. This indicates that the higher the B value of the ten nonsteroidal anti-inflammatory drugs (NSAIDs), the greater their LogK value adsorbed by the original PE microplastics. d The larger the value, the better. Model Q 2 LOO The value is 0.759 (>0.5), confirming that the model has good robustness.

[0102] Example 3: PE-UV QSAR model established according to the method in Example 1

[0103] Set pKa, LogKow, E, S, A, B, V, J - As the independent variable, PEUV-LogK d Stepwise regression analysis was performed using B as the dependent variable (specifically, the backward regression method). After automatic model identification, B and J were finally retained. - There are 2 items in the model.

[0104] In Model 1, R 2 =0.942, predictors are pKa, LogKow, E, S, B, A, V, J - However, none of the variables passed the significance test (all greater than 0.05), and except for pKa and LogKow, E, S, B, A, V, and J... - None of the variables passed the multicollinearity diagnostic test, indicating that there is a multicollinearity problem among the model parameters (VIF>10);

[0105] Furthermore, in Model 2, variables A and R are removed by going back. 2 =0.941, predictor variable is J - LogKow, pKa, E, B, V, S, but none of the variables passed the significance test (all greater than 0.05), and except for pKa and LogKow, E, S, B, V, J - None of the variables passed the collinearity diagnostic test (VIF>10);

[0106] Furthermore, in Model 3, the variable pKa and R are removed backwards. 2 =0.936, predictor variable is J - LogKow, E, B, V, S, but except J -Apart from B, none of the variables passed the significance test (all greater than 0.05), and except for LogKow, E, S, B, V, and J... - None of the variables passed the collinearity diagnostic test (VIF>10);

[0107] Furthermore, in Model 4, variables V and R are removed by going back. 2 =0.834, predictor variable is J - E, B, LogKow, S, except J - Except for B, none of the variables passed the significance test (all greater than 0.05), and except for LogKow, none of the variables passed the collinearity diagnostic test (VIF>10);

[0108] Furthermore, in Model 5, the variable LogKow is removed after stepping back. 2 =0.803, predictor variable is J - E, B, S, variable J - Variable B passed the significance test (both were 0.003 < 0.05), while variables E and S did not pass the significance test (both were greater than 0.05), and all variables failed the collinearity diagnostic test (VIF > 10).

[0109] Furthermore, in Model 6, variables S and R are removed by going back. 2 =0.748, predictor variable is J - E, B, variable J - Variable B passed the significance test (0.007 and 0.009 respectively), while variable E did not pass the significance test. All three variables passed the collinearity diagnostic test (VIF<10).

[0110] Furthermore, in Model 7, variables E and R are removed by going back. 2 =0.743, predictor variable is J - B, all variables passed the significance test (all less than 0.05) and the collinearity diagnostic test (VIF = 8.473 < 10);

[0111] Summary of Stepwise Optimization of PE-UV QSAR Model

[0112]

[0113] a. Predictor variables: (constant), J - ,A,LogKow,pKa,E,B,V,S

[0114] b. Predictor variables: (constant), J - ,LogKow,pKa,E,B,V,S

[0115] c. Predictor variables: (constant), J- ,LogKow,E,B,V,S

[0116] d. Predictor variables: (constant), J - ,LogKow,E,B,S

[0117] e. Predictor variables: (constant), J - E,B,S

[0118] f. Predictor variables: (constant), J - E,B

[0119] g. Predictor variables: (constant), J - B

[0120] h. Dependent variable: PEUV-LogK d

[0121] coefficient a

[0122]

[0123]

[0124] A significance level < 0.05 indicates that a significant influence relationship exists.

[0125] Reaction mechanism analysis

[0126] Set pKa, LogKow, E, S, A, B, V, J - As the independent variable, PEUV-LogK d Stepwise regression analysis was performed using B as the dependent variable (specifically, the backward regression method). After automatic model identification, B and J were finally retained. - With two terms in the model, the optimal QSAR model constructed is as follows:

[0127] LogK d =0.431B - 0.990J - +1.922

[0128] A detailed analysis of the model output reveals that: In the formula, the descriptor J... - The regression coefficient was -0.990 (t = -4.492, p = 0.003 < 0.01), with the largest absolute t value and the highest significance, indicating that J... - The most significant factor affecting the adsorption capacity of PE-UV microplastics for nonsteroidal anti-inflammatory drugs (NSAIDs) is the anionic coulombic interaction property of NSAIDs, which is negatively correlated with their adsorption capacity on PE-UV. The regression coefficient of B is 0.431 (t = 4.127, p = 0.004 < 0.01), indicating that B significantly affects the adsorption capacity of PE-UV microplastics on the LogK-coulombic surface.d This indicates a significant positive impact relationship between J and J. - NSAIDs with small values ​​and large B values ​​have LogK d The value will also be larger, meaning it is more easily adsorbed onto PE-UV microplastics, indicating that electrostatic interactions and hydrogen bonding are the dominant mechanisms affecting the adsorption of NSAIDs on PE-UV microplastics. Model Q 2 LOO The value is 0.578 (>0.5), confirming that the model is good.

[0129] Parameter table of the established optimal model

[0130]

[0131] Parameter table of the established optimal model

[0132]

[0133] Note: Dependent variable = PEUV - LogK d

[0134] *p<0.05, **p<0.01

[0135] Example 4: PE-HT QSAR model established according to the method in Example 1

[0136] Through multiple stepwise regression analysis, and after automatic model identification, two descriptors (B, J) were finally selected. - A QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PE-HT microplastics was constructed.

[0137] In Model 1, R 2 =0.950, predictor variable is J - A, LogKow, pKa, B, V, but except J - All variables except B failed the significance test (all greater than 0.05), and except for pKa, LogKow, and A, J... - The fact that none of the variables B and V passed the multicollinearity diagnosis indicates that there is a multicollinearity problem among the parameters of the model (VIF>10);

[0138] Furthermore, in Model 2, the variable pKa and R are removed backwards. 2 =0.946, predictor variable is J - A, LogKow, B, V, except for V (0.121>0.05), all variables passed the significance test (all less than 0.05), but except for A and LogKow, J - Both B and V failed the collinearity diagnostic test (VIF>10);

[0139] Furthermore, in Model 3, variables V and R are removed by going back. 2 =0.894, predictor variable is J - A, LogKow, and B: Variables A and LogKow did not pass the significance test (both greater than 0.05), while the predictor variables all passed the collinearity diagnostic test (VIF < 10).

[0140] Furthermore, in Model 4, variables A and R are removed by going back. 2 =0.834, predictor variable is J - B, LogKow, the variable LogKow did not pass the significance test (all greater than 0.05), while the predictor variables all passed the collinearity diagnostic test (VIF < 10);

[0141] Furthermore, in Model 5, the variable LogKow is removed after stepping back. 2 =0.790, predictor variable is J - B, all variables passed the significance test (all less than 0.05) and the collinearity diagnostic test (VIF < 10);

[0142] Summary of Stepwise Optimization of PE-HT QSAR Model

[0143]

[0144] a. Predictor variables: (constant), J - A,LogKow,pKa,B,V

[0145] b. Predictor variables: (constant), J - A,LogKow,B,V

[0146] c. Predictor variables: (constant), J - A,LogKow,B

[0147] d. Predictor variables: (constant), J - ,LogKow,B

[0148] e. Predictor variables: (constant), J - B

[0149] f. Dependent variable: PEHT-LogK d

[0150] coefficient a

[0151]

[0152] A significance level < 0.05 indicates that a significant influence relationship exists.

[0153] Reaction mechanism analysis

[0154] Parameter table of the established optimal model

[0155]

[0156] Note: Dependent variable = PEHT - LogK d

[0157] *p<0.05, **p<0.01

[0158] First, correlation analysis was performed between each descriptor and the adsorption partition coefficient. Highly linearly overlapping parameters E and S were manually filtered out. Then, pKa, LogKow, A, B, V, and J were used as the basis for further analysis. - As the independent variable, PEHT-LogK d As the dependent variable, through multiple stepwise regression analysis and automatic model identification, two descriptors (B, J) were ultimately selected. - A QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PE-HT microplastics was constructed, and the specific model equations are as follows:

[0159] LogK d =0.344B - 0.875J - +1.859

[0160] Descriptors B and J in the model - Both showed high significant correlation, with p-values ​​(0.005 and 0.002, respectively) less than 0.05, indicating that both descriptors are major factors affecting the adsorption capacity of PE-HT microplastics for nonsteroidal anti-inflammatory drugs; the regression coefficient of descriptor B was positive (0.344), and that of descriptor J was... - The regression coefficient was negative (-0.875), indicating that B affects the PEHT-LogK regression. d It produces a significant positive impact; conversely, J - Will be for PEHT-LogK d A significant negative relationship was observed. This indicates that the stronger the ability of a nonsteroidal anti-inflammatory drug (NSAID) to accept hydrogen bonding, the lower the intermolecular anionic coulombic force, exhibiting a LogK... d A higher value indicates a stronger adsorption capacity of PE-HT microplastics for this type of nonsteroidal anti-inflammatory drug (NSAID), suggesting that electrostatic interactions and hydrogen bonding are the dominant mechanisms influencing the adsorption of NSAIDs on PE-HT microplastics. Model Q 2 LOO The value is 0.657 (>0.5), confirming that the model has good robustness.

[0161] Example 5: The original PS QSAR model established according to the method in Example 1

[0162] Through multiple stepwise regression analysis, and after automatic model identification, the descriptor (B, J) was finally selected. - pKa)

[0163] In Model 1, R 2 =0.933, predictor variable is J - A, LogKow, pKa, E, B, V, S, but none of the variables passed the significance test (all greater than 0.05), and except for pKa and LogKow, E, S, B, A, V, J - None of the variables passed the multicollinearity diagnostic test, indicating that there is a multicollinearity problem among the model parameters (VIF>10);

[0164] Furthermore, in Model 2, the variable LogKow is removed after backwards. 2 =0.931, predictor variable is J - A, pKa, E, B, V, S, but none of the variables passed the significance test (all greater than 0.05), and except for pKa, none of the other variables passed the collinearity diagnostic test (VIF>10);

[0165] Furthermore, in Model 3, variables V and R are removed by going back. 2 =0.927, predictor variable is J - ,A,pKa,E,B,S,except J - All variables except B failed the significance test (all greater than 0.05), and all variables except A and pKa failed the collinearity diagnostic test (VIF>10).

[0166] Furthermore, in Model 4, variables E and R are removed by going back. 2 =0.885, predictor variable is J - ,A,pKa,B,S,except J - All variables except B failed the significance test (all greater than 0.05), while all variables except B passed the collinearity diagnostic test (VIF < 10).

[0167] Furthermore, in Model 5, variables A and R are removed by going back. 2 =0.874, predictor variable is J - pKa, B, S, variable J - B passed the significance test, while variables pKa and S did not (both greater than 0.05). Except for B, all variables passed the collinearity diagnostic test (VIF < 10).

[0168] Furthermore, in Model 6, variables S and R are removed by going back. 2 =0.813, predictor variable is J -pKa, B, except for pKa, J - Both B and B passed the significance test (both less than 0.05), and all predictor variables passed the collinearity diagnostic test (VIF < 10).

[0169] PS QSAR Model Stepwise Optimization Summary

[0170]

[0171] a. Predictor variables: (constant), J - ,A,LogKow,pKa,E,B,V,S

[0172] b. Predictor variables: (constant), J - ,A,pKa,E,B,V,S

[0173] c. Predictor variables: (constant), J - ,A,pKa,E,B,S

[0174] d. Predictor variables: (constant), J - ,A,pKa,B,S

[0175] e. Predictor variables: (constant), J - ,pKa,B,S

[0176] f. Predictor variables: (constant), J - ,pKa,B

[0177] g. Dependent variable: PS-LogK d

[0178] coefficient a

[0179]

[0180]

[0181]

[0182] A significance level < 0.05 indicates that a significant influence relationship exists.

[0183] Reaction mechanism analysis

[0184] Parameter table of the established optimal model

[0185]

[0186] Note: Dependent variable = PS - LogK d

[0187] *p<0.05, **p<0.01

[0188] Set pKa, LogKow, E, S, A, B, V, J - PS-LogK is the independent variable. d As the dependent variable, multiple regression analysis was performed, and after automatic model identification, the descriptor (B, J) was finally selected. - A QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by original PS microplastics was constructed using pKa. 2 The value is 0.813. The optimal QSAR model constructed is as follows:

[0189] LogK d =0.415B - 0.837J - -0.105pKa +1.939

[0190] Analysis shows that the electrostatic Coulomb force (J) - ) is the most important descriptor in the model. J - The regression coefficient was -0.837 (t = -4.245, p = 0.005 < 0.05), and the descriptor J... - It provides parameters for the anionic electrostatic Coulomb force of compound molecules. J - This indicates that NSAIDs weaken their adsorption tendency through electrostatic repulsion between their negatively charged surfaces and the negatively charged PS microplastics. The regression coefficient of B is 0.415 (t = -4.143, p = 0.006 < 0.05), and descriptor B is a parameter representing the strength of the hydrogen bond basicity of the compound molecule. In the model, J... - The preceding coefficient is negative and B is positive, indicating that the LogK of nonsteroidal anti-inflammatory drugs (NSAIDs) with weak anionic coulombic forces and high hydrogen bond basicity are... d A higher value indicates a greater likelihood of adsorption by PS microplastics. In short, similar to PE, nonsteroidal anti-inflammatory drugs (NSAIDs) can be adsorbed by PS microplastics through electrostatic interactions, hydrogen bonding, and other intermolecular electronic interactions. The regression coefficient for the descriptor pKa was -0.105 (t = -1.991, p = 0.094 > 0.05), which, although not passing the significance test, indicates that the pKa value affects the degree of drug dissociation and has a significant impact on parameters B and J. - It has a supplementary role. Electrostatic interactions and hydrogen bonding are the dominant mechanisms affecting the adsorption of NSAIDs on pristine PS microplastics, according to model Q. 2 LOO The value is 0.642 (>0.5), confirming that the model has good robustness.

[0191] Example 6: PS-UV QSAR model established according to the method of Example 1

[0192] Through multiple stepwise regression analysis and automatic model identification, three descriptors (A, B, V) were finally selected to construct a QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PS-UV microplastics.

[0193] In Model 1, R 2 =0.710, the predictor variables are V, A, LogKow, E, B, but none of the variables passed the significance test (all greater than 0.05), except for V, all other variables passed the collinearity diagnosis (VIF<10);

[0194] Furthermore, in Model 2, variables E and R are removed by going back. 2 =0.709, the predictor variables are V, A, LogKow, B. Except for B and V, the other variables did not pass the significance test (all are greater than 0.05), and all variables except V passed the collinearity diagnosis (VIF<10).

[0195] Furthermore, in Model 3, the variable LogKow is removed after backwards. 2 =0.709, predictors are V, A, B. Except for A, V and B both passed the significance test (both less than 0.05), and all predictors passed the collinearity test (VIF < 10);

[0196] Summary of Stepwise Optimization of PS-UV QSAR Model

[0197]

[0198] a. Predictor variables: (constant), V, A, LogKow, E, B

[0199] b. Predictors: (constant), V, A, LogKow, B

[0200] c. Predictor variables: (constant), V, A, B

[0201] d. Dependent variable: PSUV-LogK d

[0202] coefficient a

[0203]

[0204]

[0205] A significance level < 0.05 indicates that a significant influence relationship exists.

[0206] Reaction mechanism analysis

[0207] Parameter table of the established optimal model

[0208]

[0209] Note: Dependent variable = PSUV - LogK d

[0210] *p<0.05, **p<0.01

[0211] First, correlation analysis was performed between each descriptor and the adsorption partition coefficient. Highly linearly coincident parameters pKa and S were manually filtered out. Then, LogKow, E, A, B, V, and J were analyzed. - As the independent variable, PSUV-LogK d As the dependent variable, through multiple stepwise back regression analysis and automatic model identification, three descriptors (A, B, V) were finally selected to construct a QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PS-UV microplastics. The specific model equation is as follows:

[0212] LogK d = 0.372B - 0.556V - 1.579A + 0.863

[0213] In the model, descriptor A exhibited a random correlation, meaning that descriptor A did not pass the significance test (p = 0.099 > 0.05), indicating that A does not affect the PSUV-LogK correlation. d The regression coefficient of B is 0.372 (t = 3.189, p = 0.019 < 0.05), indicating that B will influence the PSUV-LogK regression relationship. d It has a positive effect. Secondly, the descriptor V is the most important factor affecting the adsorption capacity of PS-UV microplastics for nonsteroidal anti-inflammatory drugs. The regression coefficient of V is -0.556 (t=-3.714, p=0.010<0.05), indicating that V will have a positive effect on PSUV-LogK. d This produces a significant negative impact, as the molecular volume of nonsteroidal anti-inflammatory drugs (NSAIDs) is related to their LogK value when adsorbed onto the surface of PS-UV microplastics. d The values ​​show a negative correlation. This can be explained by the fact that larger drug molecules can be hindered from adsorption by steric hindrance, especially when the surface of microplastics has limited active sites, making it difficult for larger molecules to effectively contact the adsorption sites. Model Q 2 LOO The value is 0.631 (>0.5), confirming that the model has good robustness.

[0214] Example 7: PS-HT QSAR model established according to the method of Example 1

[0215] Through multiple stepwise regression analysis and automatic model identification, a QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PS-HT microplastics was finally constructed using three descriptors (A, B, V).

[0216] In Model 1, R 2 =0.896, predictor variable is J - A, LogKow, pKa, S, B, V, but none of the variables passed the significance test (all greater than 0.05). Except for B and V, all other variables passed the collinearity diagnosis (VIF < 10).

[0217] Furthermore, in Model 2, the variable J is deleted backwards. - R 2 =0.896, the predictor variables are A, LogKow, pKa, S, B, V. Except for B and V, none of the variables passed the significance test (all greater than 0.05). Except for B and V, all other variables passed the collinearity diagnosis (VIF<10).

[0218] Furthermore, in Model 3, variables S and R are removed by going back. 2 =0.893, predictor variables are A, LogKow, pKa, B, V. Except for A, B, and V, none of the variables passed the significance test (all greater than 0.05). Except for B and V, all other variables passed the collinearity diagnosis (VIF < 10).

[0219] Furthermore, in Model 4, the variable LogKow is removed after backwards. 2 =0.880, the predictor variables are A, pKa, B, V, all variables except pKa passed the significance test (all less than 0.05), and all variables except B passed the collinearity diagnosis (VIF<10);

[0220] Furthermore, in Model 5, the variable pKa and R are removed backwards. 2 =0.851, the predictor variables are A, V, B, and all variables passed the significance test (all less than 0.05) and the collinearity diagnostic test (VIF < 10);

[0221] Model Summary f

[0222]

[0223] a. Predictor variables: (constant), J - ,A,LogKow,pKa,S,B,V

[0224] b. Predictor variables: (constant), A, LogKow, pKa, S, B, V

[0225] c. Predictors: (constant), A, LogKow, pKa, B, V

[0226] d. Predictor variables: (constant), A, pKa, B, V

[0227] e. Predictor variables: (constant), A, B, V

[0228] f. Dependent variable: PSHT-LogK d

[0229] coefficient a

[0230]

[0231]

[0232] A significance level < 0.05 indicates that a significant influence relationship exists.

[0233] Reaction mechanism analysis

[0234] Parameter table of the established optimal model

[0235]

[0236] Parameter table of the established optimal model

[0237]

[0238] Note: Dependent variable = PSHT - LogK d

[0239] *p<0.05, **p<0.01

[0240] First, the highest linear overlap variable E was manually removed by performing correlation analysis on each descriptor. Then, pKa, LogKow, S, A, B, V, J were used to filter out the variables. - As the independent variable, PSHT-LogK d As the dependent variable, through multiple stepwise back regression analysis and automatic model identification, three descriptors (A, B, V) were finally selected to construct a QSAR model for the adsorption of nonsteroidal anti-inflammatory drugs by PS-HT microplastics. The specific model equation is as follows:

[0241] LogK d = 0.508B - 0.697V - 1.786A + 0.566

[0242] Analysis based on the model fitting results: The regression coefficient of A is -1.786 (t = -2.776, p = 0.032 < 0.05), indicating that A will affect the PSHT-LogK regression. d It produces a significant negative impact. The regression coefficient of B is 0.508 (t = 5.487, p = 0.002 < 0.01), indicating that B has a significant negative impact on PSHT-LogK. dA significant positive influence is observed. Descriptors A and B describe specific interactions related to intermolecular hydrogen bonds. NSAIDs, acting as hydrogen bond acceptors, can form stable hydrogen bonds with hydrogen bond donor groups on the microplastic surface, driving adsorption through chemical interactions. PS-HT microplastics have more oxygen-containing functional groups (such as carboxyl and hydroxyl groups, hydrogen bond donor groups) on their surface compared to original PS, thus exhibiting a larger adsorption coefficient for NSAIDs. V is the most important descriptor affecting the adsorption capacity of microplastics for nonsteroidal anti-inflammatory drugs (NSAIDs). The regression coefficient of V is -0.697 (t = -5.857, p = 0.001 < 0.01), indicating that V significantly influences the adsorption capacity of PSHT-LogK. d A significant negative relationship is generated; the molecular size of nonsteroidal anti-inflammatory drugs (NSAIDs) is related to PSHT-LogK. d The values ​​are negatively correlated, indicating that NSAIDs can be adsorbed by PS-HT microplastics through physical interactions involving surface distribution. Model Q 2 LOO The value is 0.712 (>0.5), confirming that the model has good robustness.

[0243] Comparative analysis results of six QSAR models

[0244] Table of QSAR model fitting equations for the adsorption of nonsteroidal anti-inflammatory drugs by six microplastics

[0245]

[0246]

[0247]

[0248] From the overall equation table of the constructed QSAR models, it can be seen that the R-values ​​of the six models are... 2 The range of values ​​(0.709–0.856) indicates that the model has a high fitting ability. The partitioning of nonsteroidal anti-inflammatory drugs (NSAIDs) between microplastics and the aqueous phase is largely influenced by the drug dissociation constant (A), the excess molecular molar refractive index (B), the molecular volume (V), and the electrostatic coulombic interaction (J). - Due to the influence of ), all six models in this study used variable B, and J was used four times. - The variables were selected twice (V variable) and once (A variable). These four molecular descriptors were introduced into different models, enabling the models to well explain the adsorption mechanism for various structurally diverse nonsteroidal anti-inflammatory drugs.

[0249] The equations for PE, PE-UV, and PE-HT show that the QSAR models for NSAID adsorption by the three PE microplastics exhibit some similarity. All three models include parameters B and J. -Where the coefficient B is positive, J - The coefficients are negative. B describes the ability of the drug molecule solute to act as a hydrogen bond acceptor, i.e., the ability of the solute molecule to accept hydrogen atoms to form hydrogen bonds. J is a parameter describing the electrostatic interaction between microplastics and organic pollutants. For the ten nonsteroidal anti-inflammatory drugs in the three models, the larger the B value, the higher the J value. - The smaller the value, the lower its LogK d The higher the value, the better. In short, the equilibrium partition coefficients of ten nonsteroidal anti-inflammatory drugs between PE microplastics and the aqueous phase before and after aging are mainly affected by intermolecular hydrogen bonding and electrostatic interactions.

[0250] For the model equations of the three PS microplastics, due to the differences in the types of compounds and microplastics included in the models, the molecular descriptor parameters with statistical significance in each model also show significant differences. The molecular descriptor used in the original QSAR model for NSAID adsorption of PS is consistent with that of the three PE models, all including parameters B and J. - And also, the coefficient of B is positive, J - Negative coefficients indicate that intermolecular hydrogen bonding and electrostatic interactions primarily influence the adsorption of NSAIDs on the original PS microplastic surface. Both the PS-UV and PS-HT QSAR models for NSAID adsorption include three parameters: A, B, and V (B is positive, A and V are negative). A represents the solute's ability to act as a hydrogen bond donor, i.e., the ability of the solute molecule to provide hydrogen atoms to form hydrogen bonds with other molecules. B represents the solute's ability to act as a hydrogen bond donor; when acceptor groups capable of forming hydrogen bonds with NSAIDs exist on the microplastic surface, the stronger the NSAID's hydrogen bond donor ability, the easier it is for it to be adsorbed by the microplastic through hydrogen bonding. V represents the molar volume of each drug molecule, reflecting the size of the solute molecule and steric hindrance. Drugs with larger molar volumes may experience greater steric hindrance when approaching the microplastic surface, hindering adsorption; while smaller NSAIDs are more likely to reach the adsorption sites on the microplastic surface. The entire process is a physisorption process primarily driven by partitioning. For the ten nonsteroidal anti-inflammatory drugs in the three models, the larger the B value, the smaller the A and V values, and the higher their LogK values. d The larger the value, the better. This indicates the equilibrium partition coefficients of ten nonsteroidal anti-inflammatory drugs between PS microplastics and the aqueous phase before and after aging, which are mainly affected by electrostatic interactions, hydrogen bonding, and partitioning. Figure 1 As shown: Six models LogK d The predicted and experimental values ​​showed a good linear correlation, and the data points were evenly distributed on both sides of the baseline, indicating that the model's systematic error was small. The model's coefficient of determination R was quantitatively evaluated. 2 Corrected determination coefficient R 2 adjAnd the root mean square error (RMSE), these metrics fully demonstrate that the six models have excellent fitting performance; such as Figure 2 As shown in the Williams application domain diagram, we can see that: 1. The standard deviations of the six models are within the normal range and there are no outliers; 2. The predicted values ​​of each compound are all within the application domain and do not exceed the warning threshold.

Claims

1. A method for constructing a QSAR model to predict the adsorption of organic pollutants by microplastics, characterized in that, Includes the following steps: 1) Obtain the partition coefficient K of organic pollutants in the microplastic phase. d value; 2) Obtain chemical descriptors for organic pollutants, including: octanol-water partition coefficient LogKow, dissociation constant pKa, excess molecular molar refractive index E, molecular dipole / polarizability S, hydrogen bond acidity A, hydrogen bond basicity B, molecular volume V, and electrostatic Coulomb force parameter J. - ; 3) Take the K obtained in step 1) d The values ​​are used as the dependent variable, and the chemical descriptors obtained in step 2) are used as the independent variables. A stepwise multiple linear regression is used to construct a QSAR model.

2. The construction method according to claim 1, characterized in that, Step 1) specifically involves: 1.1 The equilibrium adsorption capacity q was obtained through isothermal adsorption experiments. e , 1.2 Obtain K through further calculation d Value, the K d The specific formula for calculating the value is as follows: Among them, K d The unit is L / Kg; q e The equilibrium adsorption capacity of organic pollutants in the microplastic phase is expressed in μmol / Kg. C0 represents the initial concentration of organic pollutants, expressed in μmol / L. C e This represents the equilibrium concentration of organic pollutants in the aqueous phase, expressed in μmol / L. V is the volume of the aqueous solution in the adsorption experiment, in L.

3. The construction method according to claim 1 or 2, characterized in that, The model building method in step 3) is specifically to use the stepwise regression method in the multiple linear regression (MLR) algorithm of SPSS 23.0 software to build the model. Based on the statistical significance criterion, predictor variables are introduced or removed stepwise. In each iteration, the algorithm selects the optimal combination of variables that minimizes the sum of squared residuals. By stepwise optimization, the explanatory power and prediction accuracy of the model are improved, and the QSAR model is obtained.

4. The construction method according to claim 3, characterized in that, Step 3) yields the formula for the QSAR model: LogK d =N1Z1+N2Z2+N3Z3+N4Z4+N5Z5+N6Z6+N7Z7+N8Z8+C; where Z i N is one of the eight descriptor arguments mentioned in step 2); i C is the coefficient; C is the constant term.

5. A QSAR model for predicting the adsorption of organic pollutants by microplastics, based on the construction method described in claim 3, characterized in that, The formula for the QSAR model is: LogK d =N1Z1+N2Z2+N3Z3+N4Z4+N5Z5+N6Z6+N7Z7+N8Z8+C; where Z i N is one of the eight descriptor arguments mentioned in step 2); i C is the coefficient; C is the constant term.

6. The QSAR model according to claim 5, characterized in that, The QSAR model specifically includes the following formula: when the organic pollutant is a nonsteroidal anti-inflammatory drug, and, When the microplastic is a raw PE microplastic, the QSAR model formula (1) is: LogK d =0.450B-1.039J - +1.828(1) When the microplastic is a PE-UV microplastic, the QSAR model formula (2) is: LogK d =0.431B-0.990J - +1.922(2) When the microplastic is a PE-HT microplastic, the QSAR model formula (3) is: LogK d =0.344B-0.875J - +1.859(3) When the microplastic is a raw PS microplastic, the QSAR model - formula (4) is: LogK d =0.415B-0.837J - -0.105pKa+1.939(4) When the microplastic is a PS-UV microplastic, the QSAR model is given by formula (5): LogK d =0.372B-0.556V-1.579A+0.863(5) When the microplastic is a PS-HT microplastic, the QSAR model is given by formula (6): LogK d =0.508B-0.697V-1.786A+0.566(6)。 7. The method for verifying and evaluating the QSAR model formula according to claim 5, characterized in that, The QSAR model was validated using the following values, including the squared correlation coefficient R. 2 Adjusted squared correlation coefficient R 2 adj Fisher's F-test, significance test p-value, root mean square error (RMSE), variation inflation factor (VIF), and Durbin-Watson cross-validation (DW) were used to assess the robustness of the QSAR model. 2 LOO The applicability of the QSAR model was evaluated using Williams plots.

8. The construction method according to any one of claims 3 and 4 is used for the prediction of microplastic adsorption of organic pollutants.

9. The application of the QSAR model according to claim 5 in predicting the adsorption of organic pollutants by microplastics.

10. The application of the QSAR model according to claim 6 in predicting the adsorption of nonsteroidal anti-inflammatory drugs by microplastics.

Images