Predictive QSAR Models for the Toxicity of Disinfection Byproducts

Multi-Endpoint Acute Toxicity Assessment of Organic Compounds Using Large-Scale Machine Learning Modeling

In recent years, alternative animal testing methods such as computational and machine learning approaches have become increasingly crucial for toxicity testing. However, the complexity and scarcity of available biomedical data challenge the development of predictive models. Combining nonlinear machine learning together with multicondition descriptors offers a solution for using data from various assays to create a robust model. This work applies multicondition descriptors (MCDs) to develop a QSTR (Quantitative Structure-Toxicity Relationship) model based on a large toxicity data set comprising more than 80,000 compounds and 59 different end points (122,572 data points). The prediction capabilities of developed single-task multi-end point machine learning models as well as a novel data analysis approach with the use of Convolutional Neural Networks (CNN) are discussed. The results show that using MCDs significantly improves the model and using them with CNN-1D yields the best result (R2train = 0.93, R2ext = 0.70). Several structural features showed a high level of contribution to the toxicity, including van der Waals surface area (VSA), number of nitrogen-containing fragments (nN+), presence of S-P fragments, ionization potential, and presence of C-N fragments. The developed models can be very useful tools to predict the toxicity of various compounds under different conditions, enabling quick toxicity assessment of new compounds.

Photodegradation of polycyclic aromatic hydrocarbons on soil surface: Kinetics and quantitative structure-activity relationship (QSAR) model development

Polycyclic aromatic hydrocarbons (PAHs) have attracted much attention because of their widespread existence and toxicity. Photodegradation is the main natural decay process of PAHs in soil. The photodegradation kinetics of benzopyrene (BaP) on 16 kinds of soils and 10 kinds of PAHs on Hebei (HE) soil were studied. The results showed that BaP had the highest degradation rate in Shaanxi (SN) soil (kobs = 0.11 min-1), and anthracene (Ant) was almost completely degraded after 16 h of irradiation in HE soil. Two quantitative structure-activity relationship (QSAR) models were established by the multiple linear regression (MLR) method. The developed QSAR models have good stability, robustness and predictability. The model revealed that the main factors affecting the photodegradation of PAHs are soil organic matter (SOM) and the energy gap between the highest occupied molecular orbital and the lowest unoccupied molecular orbital (Egap). SOM can function as a photosensitizer to induce the production of active species for photodegradation, thus favoring the photodegradation of PAHs. In addition, compounds with lower Egap are less stable and more reactive, and thus are more prone to photodegradation. Finally, the QSAR model was optimized using machine learning approach. The results of this study provide basic information on the photodegradation of PAHs and have important significance for predicting the environmental behavior of PAHs in soil.

Application of Machine Learning Methods to Predict the Air Half-Lives of Persistent Organic Pollutants

Persistent organic pollutants (POPs) are ubiquitous and bioaccumulative, posing potential and long-term threats to human health and the ecological environment. Quantitative structure-activity relationship (QSAR) studies play a guiding role in analyzing the toxicity and environmental fate of different organic pollutants. In the current work, five molecular descriptors are utilized to construct QSAR models for predicting the mean and maximum air half-lives of POPs, including specifically the energy of the highest occupied molecular orbital (HOMO_Energy_DMol3), a component of the dipole moment along the z-axis (Dipole_Z), fragment contribution to SAscore (SAscore_Fragments), subgraph counts (SC_3_P), and structural information content (SIC). The QSAR models were achieved through the application of three machine learning methods: partial least squares (PLS), multiple linear regression (MLR), and genetic function approximation (GFA). The determination coefficients (R2) and relative errors (RE) for the mean air half-life of each model are 0.916 and 3.489% (PLS), 0.939 and 5.048% (MLR), 0.938 and 5.131% (GFA), respectively. Similarly, the determination coefficients (R2) and RE for the maximum air half-life of each model are 0.915 and 5.629% (PLS), 0.940 and 10.090% (MLR), 0.939 and 11.172% (GFA), respectively. Furthermore, the mechanisms that elucidate the significant factors impacting the air half-lives of POPs have been explored. The three regression models show good predictive and extrapolation abilities for POPs within the application domain.

A comprehensive evaluation of liposome/water partition coefficient prediction models based on the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) method: Challenges from different descriptor dimension reduction methods and machine learning algorithms

The liposome/water partition coefficient (K_lip/w) is a key parameter to evaluate the bioaccumulation potential of pollutants. Considering that it is difficult to determine the K_lip/w values of all pollutants through experiments, researchers gradually developed models to predict it. However, there is currently no research on how to comprehensively evaluate prediction models and recommend a compelling optimal modeling method. To remedy the defect of single parameters in a traditional model comparison, the TOPSIS evaluation method, based on entropy weight, was first proposed. We use this method to comprehensively evaluate models from multiple angles in this study. Thirty QSPR models, including 3 descriptor dimension reduction methods and 10 algorithms (belonging to 4 tribes), were used to predict K_lip/w and verify the effectiveness of the comprehensive assessment method. The results showed that RF (descriptor dimension reduction method), symbolism (tribes) and RF (algorithm) exhibited significant advantages in establishing the K_lip/w value prediction model. At present, the application of TOPSIS in environmental model evaluations is almost absent. We hope that the proposed TOPSIS evaluation method can be applied to more chemical datasets and provide a more systematic and comprehensive basis for the application of the QSPR model in environmental studies and other fields.

Predicting oxidative stress induced by organic chemicals by using quantitative Structure-Activity relationship methods

Cellular exposure to xenobiotic human-made products will lead to oxidative stress that gives rise to DNA damage, as well as chemical or mechanical damage. Distinguishing the chemicals that will induce oxidative stress and predicting their toxicity is necessary. In the present study, 4270 compounds in the ARE-bla assay were investigated to predict active and inactive compounds by using simple algorithms, namely, recursive partitioning (RP) and binomial logistic regression, and to develop the quantitative structure-activity relationship (QSAR) models of chemicals that activate the ARE pathway to induce oxidative stress and exert toxic effects on cells. A decision tree based on scaffold-based fragments obtained through RP analysis showed the best identification accuracy. However, the overall identification accuracy of this model for active compounds was unsatisfactory due to limited fragments. Furthermore, a binomial logistic regression model was developed from 638 active compounds and 3632 inactive chemicals. The model with a cutoff of 0.15 could predict chemicals that were active or inactive with the prediction accuracy of 69.1%. Its area under the receiver operating characteristic (ROC) curve metric (AUROC) was 0.762, which indicated the acceptable predictive ability of this model. The parameters nBM (number of multiple bonds) and H% (percentage of H atom) played dominant roles in the prediction of the activity (inactive or active) of chemicals. A global QSAR model was developed to predict the toxicity of active chemicals. However, the model displayed an unsatisfactory result with R² = 0.316 and R²_ext = 0.090. Active chemicals were then classified on the basis of structure. A total of 79 compounds with carbon chains could be predicted with acceptable performance by using a QSAR model with six descriptors (R² = 0.722, R²_ext = 0.798, Q²_Loo = 0.654, Q²_Boot = 0.755, Q²_ext = 0.721). The simple models established here contribute to efforts on identification compounds inducing oxidative stress and provide the scientific basis for risk assessment to organisms in the environment.

QSAR study of N-substituted oseltamivir derivatives as potent avian influenza virus H5N1 inhibitors using quantum chemical descriptors and statistical methods

In silico modelling studies were executed on thirty two N-substituted oseltamivir derivatives as inhibitors of influenza virus H5N1.

Emerging Polar Phenolic Disinfection Byproducts Are High-Affinity Human Transthyretin Disruptors: An in Vitro and in Silico Study

Phenolic disinfection byproducts (phenolic-DBPs) have been identified in recent years. However, the toxicity data for phenolic-DBPs are scarce, hampering their risk assessment and the development of regulations on the acceptable concentration of phenolic-DBPs in water. In this study, the binding potency and underlying interaction mechanism between human transthyretin (hTTR) and five groups of representative phenolic-DBPs (2,4,6-trihalo-phenols, 2,6-dihalo-4-nitrophenols, 3,5-dihalo-4-hydroxybenzaldehydes, 3,5-dihalo-4-hydroxybenzoic acids, halo-salicylic acids) were determined and probed by competitive fluorescence displacement assay integrated with in silico methods. Experimental results implied that 2,4,6-trihalo-phenols, 2,6-dihalo-4-nitrophenols, and 3,5-dihalo-4-hydroxybenzaldehydes have a high binding affinity with hTTR. The hTTR binding potency of the chemicals with electron-withdrawing groups on their molecular structures were higher than that with electron-donor groups. Molecular modeling methods were used to decipher the binding mechanism between model compounds and hTTR. The results documented that ionic pair, hydrogen bonding and hydrophobic interactions were dominant interactions. Finally, a mechanism-based model for predicting the hTTR binding affinity was developed. The determination coefficient ( R²), leave-one-out cross validation Q² ( Q_LOO²), bootstrapping coefficient ( Q_BOOT²), external validation coefficient ( Q_EXT²) and concordance correlation coefficient ( CCC) of the developed model met the acceptable criteria ( Q² > 0.600, R² > 0.700, CCC > 0.850), implying that the model had good goodness-of-fit, robustness, and external prediction performances. All the results indicated that the phenolic-DBPs have the hTTR disrupting effects, and further studies are needed to investigate their other mechanism of endocrine disruption.

Predicting the cytotoxicity of disinfection by-products to Chinese hamster ovary by using linear quantitative structure-activity relationship models

A suitable model to predict the toxicity of current and continuously emerging disinfection by-products (DBPs) is needed. This study aims to establish a reliable model for predicting the cytotoxicity of DBPs to Chinese hamster ovary (CHO) cells. We collected the CHO cytotoxicity data of 74 DBPs as the endpoint to build linear quantitative structure-activity relationship (QSAR) models. The linear models were developed by using multiple linear regression (MLR). The MLR models showed high performance in both internal (leave-one-out cross-validation, leave-many-out cross-validation, and bootstrapping) and external validation, indicating their satisfactory goodness of fit (R² = 0.763-0.799), robustness (Q²_LOO = 0.718-0.745), and predictive ability (CCC = 0.806-0.848). The generated QSAR models showed comparable quality on both the training and validation levels. Williams plot verified that the obtained models had wide application domains and covered the 74 structurally diverse DBPs. The molecular descriptors used in the models provided comparable information that influences the CHO cytotoxicity of DBPs. In conclusion, the linear QSAR models can be used to predict the CHO cytotoxicity of DBPs.

Transformation of norfloxacin during the chlorination of marine culture water in the presence of iodide ions

The antibacterial agent norfloxacin (NOR) and sodium hypochlorite (NaClO), which are both widely used in marine culture, react with each other to form the halogenated disinfection byproducts (X-DBPs). The effects of the water characteristics and iodide concentration on the reaction kinetics were investigated. The results showed that the reaction rate of NOR with NaClO increases from 0.0586 min^-1 to 0.1075 min^-1 when the iodide concentration was changed from 0 μg^-1 to 50 μg^-1. This demonstrated the enhancement of NOR oxidation in the presence of iodide ions. Four novel iodinated DBPs (I-DBPs) were identified in the marine culture water. Iodine substitutions occurred at the C3 and C8 positions of NOR. The formation mechanisms of X-DBPs in the marine culture water were proposed based on the intermediate and final products. NOR may undergo a ring-opening reaction, a de-carbonyl reaction and substitution to form intermediates and finally generate the X-DBPs. Furthermore, the predicted logK_OW and logBCF values of the I-DBPs were higher than that of the Br-DBPs and Cl-DBPs. The AOX concentration in the synthetic water samples decreased in the following order: seawater (8.49 mg L^-1) > marine culture water (4.05 mg L^-1) > fresh water (1.89 mg L^-1). The amount of AOX also increased with the increase in iodide concentration. These results indicated that the I-DBPs were more toxic than their brominated and chlorinated analogues.

Development of QSAR machine learning-based models to forecast the effect of substances on malignant melanoma cells

SK-MEL-5 is a human melanoma cell line that has been used in various studies to explore new therapies against melanoma in different in vitro experiments. Based on this study we report on the development of quantitative structure-activity relationship (QSAR) models able to predict the cytotoxic effect of diverse chemical compounds on this cancer cell line. The dataset of cytotoxic and inactive compounds were downloaded from the PubChem database. It contains the data for all chemical compounds for which cytotoxicity results expressed by GI₅₀ was recorded. In total 13 blocks of molecular descriptors were computed and used, after appropriate pre-processing in building QSAR models with four machine learning classifiers: Random forest (RF), gradient boosting, support vector machine and random k-nearest neighbors. Among the 186 models reported none had a positive predictive value (PPV) higher than 0.90 in both nested cross-validation and on an external dataset testing, but 7 models had a PPV higher than 0.85 in both evaluations, all seven using the RFs algorithm as a classifier, and topological descriptors, information indices, 2D-autocorrelation descriptors, P-VSA-like descriptors, and edge-adjacency descriptors as sets of features used for classification. The y-scrambling test was associated with considerably worse performance (confirming the non-random character of the models) and the applicability domain was assessed through three different methods.