Method for evaluating liquid crystal monomer mixing environment risk based on machine learning
By constructing a machine learning-based QSAR model, using mixed molecular descriptors and environmental effect factors, predicting the mixed effect factors of liquid crystal monomer mixtures in the environment, and combining Monte Carlo simulation to evaluate the toxicity equivalent of liquid crystal monomers, the problem of failure to fully consider the interactions of multiple liquid crystal monomers in the prior art is solved, and a more accurate environmental risk assessment is achieved.
Patent Information
- Application Number
- CN202411628896.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-04-11
AI Technical Summary
The existing environmental risk assessment methods for liquid crystal monomers fail to fully consider the interaction between multiple liquid crystal monomers, resulting in inaccurate evaluation results and inability to truly reflect environmental exposure risks.
By constructing a quantitative structure-effect relationship (QSAR) model based on machine learning, the mixed molecular descriptor and environmental effect factors are used to predict the mixed effect factors of liquid crystal monomer mixtures in the environment, and the toxicity equivalent of liquid crystal monomer is evaluated in combination with Monte Carlo simulations.
This method can more accurately predict the mixed environmental risks of liquid crystal monomers, considering the interaction between pollutants, providing a more realistic environmental exposure and avoiding overestimation of environmental risks.
Smart Images
Figure CN119811515A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of environmental science and technology, and particularly focuses on the application of machine learning in the environmental risk assessment of liquid crystal monomer mixing. By constructing a quantitative structure-activity relationship (QSAR) model based on molecular descriptors and mixed molecular descriptors, the present invention uses machine learning technology to predict the mixed environmental effects of liquid crystal monomers. At the same time, the present invention combines the occurrence concentration of liquid crystal monomers in the environment to evaluate their potential environmental risks. Background Art
[0002] Liquid crystal monomers are a class of artificially synthesized organic pollutants with liquid crystal properties. They are widely used in liquid crystal displays and therefore play an important role in electronic products such as computers, televisions, and smart phones. With the rapid development of electronic products, the use of liquid crystal monomers is increasing. However, since most of them are semi-volatile organic compounds and do not form chemical bonds with other substrates in liquid crystal displays, liquid crystal monomers can easily escape from liquid crystal displays and enter the environment, causing certain ecological impacts.
[0003] In recent years, the environmental science community has paid more and more attention to the environmental risks of liquid crystal monomers and their comprehensive impacts. The existing risk assessment system generally focuses on risk assessment tests of single substances under laboratory conditions, and fails to fully consider the mixing effects between multiple liquid crystal monomers. This limitation makes it difficult for traditional assessment methods to accurately reflect the real environmental risks of liquid crystal monomers and may underestimate their potential impact on the ecological environment system.
[0004] Liquid crystal monomers usually exist in the form of mixtures in the environment, which means that different liquid crystal monomers may interact with each other and form complex ecological effects. Although studies have shown the potential risks of liquid crystal monomers and the need for in-depth assessment of their mixed effects, the lack of relevant mixed environmental data, complex interactions and individual differences in different environments have increased the challenges of improving assessment methods.
[0005] Traditional quantitative structure-activity relationship (QSAR) models often rely on linear regression analysis and simple statistical techniques when they are constructed, which makes them inadequate when dealing with complex data sets and multiple variables, thus limiting the predictive accuracy of the models. In contrast, machine learning technology has shown significant advantages in building QSAR models. By using more complex algorithms, machine learning can process larger-scale, multi-dimensional data, capture nonlinear relationships and complex interactions between variables, and thus more accurately predict the mixed environmental risks of liquid crystal monomers, providing a more reliable basis for assessing their potential impact in the environment.
[0006] Current methods for assessing exposure risk of liquid crystal monomers have the following problems
[0007] (1) The assessment results ignore the interactions between different types of liquid crystal monomers (such as synergistic and antagonistic effects). Current methods usually first evaluate the exposure risk of a single liquid crystal monomer and then simply add them together to obtain the mixed exposure risk of the liquid crystal monomers. However, there are potential interactions between different pollutants. These interactions may promote or inhibit the expression of the environmental effects of liquid crystal monomers, resulting in inaccurate assessment results that cannot truly reflect environmental exposure risks. The current assessment method fails to take into account the interactions between pollutants.
[0008] (3) The assessment results are accidental and fail to reflect the actual exposure environment. When calculating the daily intake, the current assessment method usually uses the average or median concentration of liquid crystal monomers, instead of using the Monte Carlo model to simulate all possible exposure concentrations in the actual environment. This may lead to an overestimation or underestimation of the assessment results.
[0009] (4) Inaccuracy in model building. In the process of model building, the current method is too simple in screening molecular descriptors, and only deletes molecular descriptors with all zeros and high correlation. This screening method is not rigorous enough. In addition, the final model construction adopts the traditional multivariate linear regression principle, which may lead to relatively weak prediction ability of the model. Summary of the invention
[0010] In view of the deficiencies in the prior art, the present invention provides a method for assessing the mixed environmental risk of liquid crystal monomers in the environment based on machine learning. It is based on mixed molecular descriptors and environmental effect factors, with mixed molecular descriptors as independent variables and environmental effect factors as dependent variables, and a QSAR model is constructed through machine learning to predict the mixed effect factors of liquid crystal monomer mixtures in the environment. Finally, based on the concentration distribution of liquid crystal monomers in the environment, the toxic equivalent of the liquid crystal monomers in the environment is calculated, and the hazards of the liquid crystal monomers are evaluated based on the toxic equivalent.
[0011] This paper uses 23 liquid crystal monomers frequently detected in the environment as target substances. The CAS numbers of the 23 liquid crystal monomers are shown in Table 1.
[0012] Molecular descriptors are numerical information that quantifies the structure and properties of molecules and are used to represent different characteristics of molecules. The environmental effects of compounds are closely related to their structural characteristics. Molecular descriptors capture these key features, allowing QSAR models to establish the relationship between environmental effects and molecular structures. The molecular descriptors used in this paper can be directly obtained from the PaDEL-descriptor software, a total of 1444. The calculation of mixed molecular descriptors is based on molecular descriptors, and the calculation formulas are (1)-(8).
[0013] The environmental effect factor is an indicator used to quantify the impact of a particular chemical substance or product on the environment. It is usually used to assess the risks that a compound may cause in the environment, including its impact on ecosystems, human health, and biodiversity. The environmental effect factor used in this article is obtained from the ADMETlab 3.0 platform, which can predict the effect of a compound in the environment based on its smiles structure. The effect factor selected in this article is Drug-induced neurotoxicity.
[0014] Introduce specific steps:
[0015] 1. According to the CAS numbers of the 23 liquid crystal monomers provided in Table 1, the smiles structure of each liquid crystal monomer was obtained on the Pubchem platform.
[0016] 2. According to the smiles structure of each liquid crystal monomer, the corresponding environmental effect factor TEF is obtained on the ADMETlab 3.0 platform.
[0017] 3. The molecular descriptors of 23 liquid crystal monomers were obtained using the PaDEL-descriptor software, with 1444 molecular descriptors for each substance.
[0018] 4. Screening molecular descriptors. When building a model, a large number of molecular descriptors may cause the model to be overfitted and unstable, so it is necessary to eliminate redundant molecular descriptors. The screening method uses recursive feature elimination (RFE), and its code is shown in the appendix.
[0019] 5. Calculate the mixed molecular descriptor and mixed effect factor based on the screened molecular descriptor. The calculation formula is as follows (1)-(8); calculate the mixed environmental effect factor based on the environmental effect factor, the calculation formula is (9)
[0020]
[0021] Wherein, i represents the number of a single liquid crystal monomer, n represents the total number of liquid crystal monomers (n=23), D mix,i represents the mixed molecular descriptor of liquid crystal monomers obtained by different calculation formulas, x i represents the molecular descriptor of a single liquid crystal monomer remaining after screening in step 2, TEF i Represents the environmental effect factor of a single liquid crystal monomer, TEF mix,i represents the mixed environmental effect factor, p i It represents the percentage of pollutant concentration calculated by Monte Carlo model simulation. Monte Carlo simulation is realized by Crystal Ball software.
[0022] 6. Using the mixed molecular descriptors calculated above as independent variables and the mixed environmental effect factors as dependent variables, a QSAR model was constructed through machine learning (support vector machine SVM).
[0023] 7. Model performance evaluation: We used the root mean square error (RMSE) of the test set and the cross-validation coefficient of determination (R 2 ) as the key indicator. In the formula, R 2 The calculation of is as follows:
[0024]
[0025] Among them SS res is the residual sum of squares, which represents the difference between the predicted value and the true value, SS tot is the total sum of squares, which indicates the degree of variation of the true value relative to its mean. This result reflects that the model has good fitting ability.
[0026] At the same time, the root mean square error (RMSE) indicator is calculated as follows:
[0027]
[0028] In this formula, y i is the true value, is the predicted value and n is the total number of samples.
[0029] 8. Based on the QSAR model constructed in step 6, the mixed molecular descriptor calculated under the typical exposure concentration (P50) of Monte Carlo simulation is used as the independent variable and brought into the QSAR model to obtain the corresponding TEF mix .
[0030] 9. The typical exposure concentration (P50) was obtained by using the average and standard deviation of the environmental concentrations of 23 liquid crystal monomers through 10,000 normal Monte Carlo simulations.
[0031] 10. Calculate the daily intake of liquid crystal monomers in dust through hand-to-mouth ingestion, breathing inhalation, skin contact and the sum of the three pathways, as well as the environmental effect equivalent. The calculation formula is as follows:
[0032]
[0033] EDI total =EDI ing +EDI inh +EDI der #(15)
[0034] TEQ mix =EDI total ×TEF mix #(16)
[0035] EDI der 、EDI ing 、EDI inh and EDI total The daily intake of liquid crystal monomers in dust through skin contact, hand-to-mouth ingestion, respiratory inhalation and the sum of the three routes, ng / kg BW / day; C is the content of liquid crystal monomers in dust, here using the P50 after Monte Carlo simulation, ng / g; IngR is the intake rate of adults (0.02g / day) or children (0.05g / day); EF is the exposure frequency of adults (indoors = 1159 minutes / day, outdoor = 281 minutes / day) or children (indoors = 1278 minutes / day, outdoor = 107 minutes / day) in one day; BW refers to the average weight of adults (70 kg) or children (15 kg); InhR is the average weight of adults (15.7m 3 / day) or children (10.1m 3 / day) inhalation rate; PEF is the particle emission factor (1.36×10 6 m 3 / g; SA is for adults (4350cm 2 ) or children (1600cm 2 ) of the skin contact area; SAF is an adult (7×10 -5 g / cm 2 ) or children (2×10 -4 g / cm 2 ) is the dust skin adhesion factor; DAF is the skin absorption factor (0.13).
[0036] The present invention provides a method for assessing the mixed environmental risk of liquid crystal monomers in the environment. The method is different from the general risk assessment method in that:
[0037] 1. It is more reasonable to incorporate machine learning to screen molecular descriptors.
[0038] 2. Use machine learning to build models with stronger fitting capabilities.
[0039] 3. Use the Monte Carlo model to simulate all possible concentrations of liquid crystal monomers in the actual environment.
[0040] The present invention calculates the environmental effect factor TEF of each liquid crystal monomer, screens out mixed molecular descriptors, and uses machine learning to construct a QSAR model to predict the mixed TEF of 23 liquid crystal monomers under typical exposure concentrations. Finally, the environmental risk assessment is conducted based on the typical occurrence concentration of liquid crystal monomers in the actual environment, providing more accurate data support for evaluating the hazards of liquid crystal monomers. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 Flow chart for evaluating the environmental risk of liquid crystal monomers according to the present invention. DETAILED DESCRIPTION
[0042] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described examples are only part of the implementation cases of the present invention, not all the implementation cases. Based on the implementation cases of the present invention, all other implementation cases obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.
[0043] The flow chart of the method for predicting the risk of mixed neurological disease of liquid crystal monomers of the present invention is as follows: Figure 1 This implementation case evaluates the environmental risk of liquid crystal monomers actually detected in dust, where the pollutant concentration is the average concentration. Specific information is shown in Table 1.
[0044] The process of establishing the liquid crystal monomer environmental risk assessment model and the mixed molecular descriptor screening in the present invention is as follows: Figure 1 As shown, it mainly includes the following steps:
[0045] (1) According to the above steps 1-3, the final TEF of the CAS number of each liquid crystal monomer is presented in Table 1 and Table 2. The 1444 molecular descriptors of the 23 liquid crystal monomers obtained are in the Excel table provided.
[0046] (2) The three molecular descriptors selected in step 4 above are MATS2i, GATS8p and ETA_Eta_R. MATS2i is the Moran autocorrelation with a lag of 2 weighted by the first ionization potential; GATS8p is the Geary autocorrelation with a lag of 8 weighted by the polarizability; and ETA_Eta_R is the comprehensive index Eta of the reference alkane.
[0047] (3) The hybrid molecular descriptor and hybrid TEF obtained by the above step 5 according to formulas (1)-(8) and calculated by Monte Carlo simulation 10,000 times mix See Excel Table 2.
[0048] (4) The performance of the QSAR model constructed in steps 6-7 is R 2 The value is 0.6923 and the root mean square error (RMSE) of the test set is 0.0247.
[0049] (5) According to the above steps 8-10, the TEF mix is 0.52. The calculated EDI and effect equivalent TEQ are shown in Table 2 and Table 3 respectively.
[0050] The cross-validation determination coefficient of the QSAR (quantitative structure-activity relationship) model constructed using machine learning is (R 2 =0.6923), which means that the model can explain about 69.23% of the variation of the target variable, reflecting the good fit of the model to the training set, and also reflecting its strong generalization ability when processing new data. The root mean square error (RMSE) of this model is 0.0247, indicating that the average error of the model prediction is relatively small. The lower the RMSE, the closer the model's prediction to the actual observation value, reflecting the accuracy of the model in the specific application scenario. Taking these two indicators into consideration, we can conclude that this model shows strong stability and accuracy, and can effectively predict the effect factors of the target pollutants.
[0051] We used the 50% quantile (P50) as the typical exposure concentration to calculate the EDI and mixed TEF (0.52), and calculated the mixed TEQ for adults and children to be 0.088 ng / kg BW / day and 1.13 ng / kg BW / day, respectively. Compared with the traditional exposure risk assessment method (see Table 3), it was found that the traditional method overestimated the environmental risk of liquid crystal monomers in dust, indicating that the effects of multiple liquid crystal monomers on the environment may be antagonistic. The traditional assessment model cannot predict the interaction between pollutants, resulting in a failure to reflect the exposure situation in the real environment.
[0052] In summary, this method based on the mixed exposure principle and the QSAR model constructed by machine learning shows significant superiority in environmental risk assessment. It not only performs well in data fitting and prediction accuracy, but also considers the interactions between pollutants more comprehensively. Compared with traditional methods, this model can provide more realistic environmental exposure conditions, avoid overestimation of environmental risks, and thus provide a scientific basis for the formulation of more effective environmental protection policies and management measures. The introduction of this method helps to more accurately assess the impact of target pollutants, provide strong support for environmental health research and the formulation of public policies, and has important application value and social significance.
[0053] Table 1 Concentration of liquid crystal monomers in dust
[0054]
[0055] Table 2 Environmental effect parameters of liquid crystal monomers
[0056]
[0057] Table 3 Results of exposure risk assessment using traditional methods
[0058]
[0059]
Claims
1. A method for evaluating the environmental risk of liquid crystal monomer mixing based on machine learning, characterized in that: The following steps are involved: 1). According to the CAS numbers of the 23 liquid crystal monomers provided in Table 1, the smiles structure of each liquid crystal monomer was obtained on the Pubchem platform; 2).According to the smiles structure of each liquid crystal monomer, the corresponding environmental effect factor TEF is obtained on the ADMETlab 3.0 platform; 3). Obtain molecular descriptors of 23 liquid crystal monomers using PaDEL-descriptor software, with 1444 molecular descriptors for each substance; 4). Screening molecular descriptors; when building the model, eliminate redundant molecular descriptors; the screening method uses feature recursive elimination; 5). Calculate the mixed molecular descriptor and the mixed effect factor based on the screened molecular descriptor; the calculation formula is as follows (1)-(8); calculate the mixed environmental effect factor based on the environmental effect factor, the calculation formula is (9) Where i represents the number of a single liquid crystal monomer, n represents the total number of liquid crystal monomers, n = 23, D mix,i represents the mixed molecular descriptor of liquid crystal monomers obtained by different calculation formulas, x i Represents the molecular descriptor of a single liquid crystal monomer remaining after screening in step 2), TEF i Represents the environmental effect factor of a single liquid crystal monomer, TEF mix,i represents the mixed environmental effect factor, p i It represents the percentage of pollutant concentration calculated by Monte Carlo model simulation; Monte Carlo simulation is implemented by Crystal Ball software; 6.) Using the calculated mixed molecular descriptors as independent variables and the mixed environmental effect factors as dependent variables, a QSAR model was constructed through machine learning; 7) The QSAR model constructed in step 6) uses the mixed molecular descriptor calculated at the typical exposure concentration P50 of Monte Carlo simulation as the independent variable and is introduced into the QSAR model to obtain the corresponding TEF mix ; 2. The method according to claim 1, characterized in that: The typical exposure concentration P50 was obtained by using the average and standard deviation of the environmental concentrations of 23 liquid crystal monomers through 10,000 normal Monte Carlo simulations.