Prediction of Aquatic Toxicity Mode of Action Using Linear Discriminant and Random Forest Models

Chemical Mixtures and Multiple Stressors: Same but Different?

Ecosystems are strongly influenced by multiple anthropogenic stressors, including a wide range of chemicals and their mixtures. Studies on the effects of multiple stressors have largely focussed on nonchemical stressors, whereas studies on chemical mixtures have largely ignored other stressors. However, both research areas face similar challenges and require similar tools and methods to predict the joint effects of chemicals or nonchemical stressors, and frameworks to integrate multiple chemical and nonchemical stressors are missing. We provide an overview of the research paradigms, tools, and methods commonly used in multiple stressor and chemical mixture research and discuss potential domains of cross-fertilization and joint challenges. First, we compare the general paradigms of ecotoxicology and (applied) ecology to explain the historical divide. Subsequently, we compare methods and approaches for the identification of interactions, stressor characterization, and designing experiments. We suggest that both multiple stressor and chemical mixture research are too focused on interactions and would benefit from integration regarding null model selection. Stressor characterization is typically more costly for chemical mixtures. While for chemical mixtures comprehensive classification systems at suborganismal level have been developed, recent classification systems for multiple stressors account for environmental context. Both research areas suffer from rather simplified experimental designs that focus on only a limited number of stressors, chemicals, and treatments. We discuss concepts that can guide more realistic designs capturing spatiotemporal stressor dynamics. We suggest that process-based and data-driven models are particularly promising to tackle the challenge of prediction of effects of chemical mixtures and nonchemical stressors on (meta-)communities and (meta-)food webs. We propose a framework to integrate the assessment of effects for multiple stressors and chemical mixtures. Environ Toxicol Chem 2023;42:1915-1936. © 2023 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Deep mining of reported emerging contaminants in China's surface water in the past decade: Exposure, ecological effects and risk assessment

The identification and management of high-risk contaminants have raised great concern from governments. Facing the growing amount of data on the occurrence of emerging contaminants (ECs) in surface water, a deep mining and quality control strategy was developed to integrate data on all reported ECs in Chinese surface water over the past decade, and an exposure and effect database was further built. In addition, multilevel risk characterization was carried out to prioritize high-risk areas, contaminants and endpoints. A total of 1038 ECs, mainly pharmaceutical and personal care products (PPCPs) and industrial chemicals, were curated, with concentrations ranging from 0.02 pg/L to 533 µg/L. For individual risk, all the provinces had acceptable risks except for Henan, which was characterized with a medium chronic risk. Nine ECs, including 4-nonylphenol and estrone, dominated individual risks. Conversely, for multisubstance risk, 76.20% and 73.87% of aquatic organisms were affected acutely and chronically at the national level, with acute and chronic risks exceeding the safety threshold of 5% in 11 and 19 provinces, respectively. Nineteen ECs, including sitosterol and chyfluthrin, dominated the multisubstance risk. In addition, 9 MoAs mainly inducing electron transfer inhibition, neurotoxicity and narcosis toxicity are high-risk endpoints. The study revealed the ecological risk status and key risk entities of Chinese surface waters, which provided the latest data to support the control of ECs in China.

Prior Knowledge for Predictive Modeling: The Case of Acute Aquatic Toxicity

Early assessment of the potential impact of chemicals on health and the environment requires toxicological properties of the molecules. Predictive modeling is often used to estimate the property values in silico from pre-existing experimental data, which is often scarce and uncertain. One of the ways to advance the predictive modeling procedure might be the use of knowledge existing in the field. Scientific publications contain a vast amount of knowledge. However, the amount of manual work required to process the enormous volumes of information gathered in scientific articles might hinder its utilization. This work explores the opportunity of semiautomated knowledge extraction from scientific papers and investigates a few potential ways of its use for predictive modeling. The knowledge extraction and predictive modeling are applied to the field of acute aquatic toxicity. Acute aquatic toxicity is an important parameter of the safety assessment of chemicals. The extensive amount of diverse information existing in the field makes acute aquatic toxicity an attractive area for investigation of knowledge use for predictive modeling. The work demonstrates that the knowledge collection and classification procedure could be useful in hybrid modeling studies concerning the model and predictor selection, addressing data gaps, and evaluation of models’ performance.

A critical review of effect modeling for ecological risk assessment of plant protection products

A wide diversity of plant protection products (PPP) is used for crop protection leading to the contamination of soil, water, and air, which can have ecotoxicological impacts on living organisms. It is inconceivable to study the effects of each compound on each species from each compartment, experimental studies being time consuming and cost prohibitive, and animal testing having to be avoided. Therefore, numerous models are developed to assess PPP ecotoxicological effects. Our objective was to provide an overview of the modeling approaches enabling the assessment of PPP effects (including biopesticides) on the biota. Six categories of models were inventoried: (Q)SAR, DR and TKTD, population, multi-species, landscape, and mixture models. They were developed for various species (terrestrial and aquatic vertebrates and invertebrates, primary producers, micro-organisms) belonging to diverse environmental compartments, to address different goals (e.g., species sensitivity or PPP bioaccumulation assessment, ecosystem services protection). Among them, mechanistic models are increasingly recognized by EFSA for PPP regulatory risk assessment but, to date, remain not considered in notified guidance documents. The strengths and limits of the reviewed models are discussed together with improvement avenues (multigenerational effects, multiple biotic and abiotic stressors). This review also underlines a lack of model testing by means of field data and of sensitivity and uncertainty analyses. Accurate and robust modeling of PPP effects and other stressors on living organisms, from their application in the field to their functional consequences on the ecosystems at different scales of time and space, would help going toward a more sustainable management of the environment. Graphical Abstract Combination of the keyword lists composing the first bibliographic query. Columns were joined together with the logical operator AND. All keyword lists are available in Supplementary Information at https://doi.org/10.5281/zenodo.5775038 (Larras et al. 2021).

Generating accurate in silico predictions of acute aquatic toxicity for a range of organic chemicals: Towards similarity-based machine learning methods

There has been an increase in the use of non-animal approaches, such as in silico and/or in vitro methods, for assessing the risks of hazardous chemicals. A number of machine learning algorithms link molecular descriptors that interpret chemical structural properties with their biological activity. These computer-aided methods encounter several challenges, the most significant being the heterogeneity of datasets; more efficient and inclusive computational methods that are able to process large and heterogeneous chemical datasets are needed. In this context, this study verifies the utility of similarity-based machine learning methods in predicting the acute aquatic toxicity of diverse organic chemicals on Daphnia magna and Oryzias latipes. Two similarity-based methods were tested that employ a limited training dataset, most similar to a given fitting point, instead of using the entire dataset that encompasses a wide range of chemicals. The kernel-weighted local polynomial approach had a number of advantages over the distance-weighted k-nearest neighbor (k-NN) algorithm. The results highlight the importance of lipophilicity, electrophilic reactivity, molecular polarizability, and size in determining acute toxicity. The rigorous model validation ensures that this approach is an important tool for estimating toxicity in new or untested chemicals.

Integrating QSAR models predicting acute contact toxicity and mode of action profiling in honey bees (A. mellifera): Data curation using open source databases, performance testing and validation

Honey bees (Apis mellifera) provide key ecosystem services as pollinators bridging agriculture, the food chain and ecological communities, thereby ensuring food production and security. Ecological risk assessment of single Plant Protection Products (PPPs) requires an understanding of the exposure and toxicity. In silico tools such as QSAR models can play a major role for the prediction of structural, physico-chemical and pharmacokinetic properties of chemicals as well as toxicity of single and multiple chemicals. Here, the first integrative honey bee QSAR model has been developed for PPPs using EFSA's OpenFoodTox, US-EPA ECOTOX and Pesticide Properties DataBase i) to predict acute contact toxicity (LD₅₀) and ii) to profile the Mode of Action (MoA) of pesticides active substances. Three different classification-based and four regression-based models were developed and tested for their performance, thus identifying two models providing the most reliable predictions based on k-NN algorithm. The two-category QSAR model (toxic/non-toxic; n = 411) was validated using sensitivity (=0.93), specificity (=0.85), balanced accuracy (=0.90), and Matthews correlation coefficient (MCC = 0.78) as statistical parameters. The regression-based model (n = 113) was validated for its reliability and robustness (R² = 0.74; MAE = 0.52). Current study proposes the MoA profiling for 113 pesticides active substances and the first harmonised MoA classification scheme for acute contact toxicity in honey bees, including LD_50s data points from three different databases. The classification allows to further define MoAs and the target site of PPPs active substances, thus enabling regulators and scientists to refine chemical grouping and toxicity extrapolations for single chemicals and component-based mixture risk assessment of multiple chemicals. Relevant future perspectives are briefly addressed to integrate MoA, adverse outcome pathways (AOPs) and toxicokinetic information for the refinement of single-chemical/combined toxicity predictions and risk estimates at different levels of biological organization in the bee health context.

Effect of the structural factors of organic compounds on the acute toxicity toward Daphnia magna

The acute toxicity of organic compounds towards Daphina magna was subjected to QSAR analysis. The two-dimensional simplex representation of molecular structure (2D SiRMS) and the support vector machine (SVM), gradient boosting (GBM) methods were used to develop QSAR models. Adequate regression QSAR models were developed for incubation of 24 h. Their interpretation allowed us to quantitatively describe and rank the well-known toxicophores, to refine their molecular surroundings, and to distinguish the structural derivatives of the fragments that significantly contribute to the acute toxicity (LC₅₀) of organic compounds towards D. magna. Based on the results of the interpretation of the regression models, a molecular design (modification) of highly toxic compounds was performed in order to reduce their hazard. In addition, acceptable classification QSAR models were developed to reliably predict the following mode of action (MOA): specific and non-specific toxicity of organic compounds towards D. magna. When interpreting these models, we were able to determine the structural fragments and the physicochemical characteristics of molecules that are responsible for the manifestation of one of the modes of action. The on-line version of the OCHEM expert system (https://ochem.eu), HYBOT descriptors, and the random forest and SVM methods were used for a comparative QSAR investigation.

Development of QSAAR and QAAR models for predicting fish early-life stage toxicity with a focus on industrial chemicals

We developed models for predicting fish early-life stage (ELS) toxicities oriented to industrial chemicals. The training set was constructed without data from the Office of Pesticide Programs Pesticide Ecotoxicity Database, the main source for the pesticide-biased training set used in our previous work (SAR QSAR Environ. Res. 29:9, 725-742). In addition to the descriptors from the previous study, we also used water solubility to develop the new models, which were evaluated against the test set used in our previous study so that we could focus on the effects of the different training set and the additional descriptor. The statistics for the new models were hardly better than those for the previous models, which suggests, contrary to our expectations, that pesticide-biased data can successfully be used to develop models for predicting the fish ELS toxicities oriented to industrial chemicals. Acute Daphnia magna toxicity was important for the predictive QSAARs in both studies. A distance-based method for defining the applicability domains indicated that water solubility was a key indicator for detecting underestimated chemicals. The comparison of fish ELS toxicities for chemicals presented in different literatures revealed the uncertainty of the experimental data, which may lead to the low predictivity.

Mode of Action Classifications in the EnviroTox Database: Development and Implementation of a Consensus MOA Classification

Multiple mode of action (MOA) frameworks have been developed in aquatic ecotoxicology, mainly based on fish toxicity. These frameworks provide information on a key determinant of chemical toxicity, but the MOA categories and level of specificity remain unique to each of the classification schemes. The present study aimed to develop a consensus MOA assignment within EnviroTox, a curated in vivo aquatic toxicity database, based on the following MOA classification schemes: Verhaar (modified) framework, Assessment Tool for Evaluating Risk, Toxicity Estimation Software Tool, and OASIS. The MOA classifications from each scheme were first collapsed into one of 3 categories: non-specifically acting (i.e., narcosis), specifically acting, or nonclassifiable. Consensus rules were developed based on the degree of concordance among the 4 individual MOA classifications to attribute a consensus MOA to each chemical. A confidence rank was also assigned to the consensus MOA classification based on the degree of consensus. Overall, 40% of the chemicals were classified as narcotics, 17% as specifically acting, and 43% as unclassified. Sixty percent of chemicals had a medium to high consensus MOA assignment. When compared to empirical acute toxicity data, the general trend of specifically acting chemicals being more toxic is clearly observed for both fish and invertebrates but not for algae. EnviroTox is the first approach to establishing a high-level consensus across 4 computationally and structurally distinct MOA classification schemes. This consensus MOA classification provides both a transparent understanding of the variation between MOA classification schemes and an added certainty of the MOA assignment. In terms of regulatory relevance, a reliable understanding of MOA can provide information that can be useful for the prioritization (ranking) and risk assessment of chemicals. Environ Toxicol Chem 2019;38:2294-2304. © 2019 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals, Inc. on behalf of SETAC.

Ecotoxicological QSAR modeling of organic compounds against fish: Application of fragment based descriptors in feature analysis

Organic compounds (OCs) constitute an enormously large class of highly persistent and toxic chemicals widely used for various purposes throughout the world. Their increased detection in water bodies, mainly sewage treatment plants via industrial discharge, has rendered them to become a cause for ecological concern. The limited availability of experimental toxicological data has necessitated development of models that can help us identify the most hazardous and potentially toxic compounds thus prioritizing the experiments on the selected chemicals. Computational tools such as quantitative structure-activity relationship (QSAR) can be used to predict the missing data and classify the chemicals based on their acute predicted responses for existing as well as not yet synthesized chemicals. In the current study, novel, externally validated, highly robust local QSAR models for different chemical classes and moderately robust global QSAR models were developed using partial least squares (PLS) regression technique using a large dataset of 1121 OCs for the fish mortality endpoint. For feature selection, genetic algorithm along with stepwise regression was used while model validation was performed using various stringent validation criteria following the strict rules of OECD guidelines of QSAR validation. The variables included in the models were obtained from simplex representation of molecular structures (SiRMS) (Version 4.1.2.270), Dragon (Version 7.0) and PaDEL-descriptor software (Version 2.20). The final developed models were robust, externally predictive and characterized by a large chemical as well as biological domain. The predictive efficiency of the developed models was then compared with the ECOSAR tool in order to justify the applicability of the developed models in ecotoxicological predictions for organic chemicals. Better predictive efficiency of the developed QSAR models compared to the ECOSAR derived predictions signifies their applicability in early risk assessment of known as well as untested chemicals in order to design safer alternatives to the environment.

Target site model: Application of the polyparameter target lipid model to predict aquatic organism acute toxicity for various modes of action

A database of 2049 chemicals with 47 associated modes of action (MoA) was compiled from the literature. The database includes alkanes, polycyclic aromatic hydrocarbons, pesticides, inorganic, and polar compounds. Brief descriptions of some critical MoA classification groups are provided. The MoA from the 14 sources were assigned using a variety of reliable experimental and modeling techniques. Toxicity information, chemical parameters, and solubility limits were combined with the MoA label information to create the data set used for model development. The model database was used to generate linear free energy relationships for each specific MoA using multilinear regression analysis. The model uses chemical-specific Abraham solute parameters estimated from AbSolv to determine MoA-specific solvent parameters. With this procedure, critical target site concentrations are determined for each genus. Statistical analysis showed a wide range in values of the solvent parameters for the significant MoA. Environ Toxicol Chem 2019;38:222-239. © 2018 SETAC.

Development of thresholds of excess toxicity for environmental species and their application to identification of modes of acute toxic action

The acute toxicity of organic pollutants to fish, Daphnia magna, Tetrahymena pyriformis, and Vibrio fischeri was investigated. The results indicated that the Toxicity Ratio (TR) threshold of log TR =1, which has been based on the distribution of toxicity data to fish, can also be used to discriminate reactive or specifically acting compounds from baseline narcotics for Daphnia magna and Vibrio fischeri. A log TR=0.84 is proposed for Tetrahymena pyriformis following investigation of the relationships between the species sensitivity and the absolute averaged residuals (AAR) between the predicted baseline toxicity and the experimental toxicity. Less inert compounds exhibit relatively higher toxicity to the lower species (Tetrahymena pyriformis and Vibrio fischeri) than the higher species (fish and Daphnia magna). A greater number of less inert compounds with log TR greater than the thresholds was observed for Tetrahymena pyriformis and Vibrio fischeri. This may be attributed to the hydrophilic compounds which may pass more easily through cell membranes than the skin or exoskeleton of organisms and have higher bioconcentration factors in the lower species, leading to higher toxicity. Most of classes of chemical associated with excess toxicity to one species also exhibited excess toxicity to other species, however, a few classes with excess toxicity to one species exhibiting narcotic toxicity to other species and thus may have different MOAs between species. Some ionizable compounds have log TR much lower than one because of the over-estimated log K_OW. The factors that influence the toxicity ratio calculated from baseline level are discussed in this paper.

Developing predictive models for toxicity of organic chemicals to green algae based on mode of action

Organic chemicals in the aquatic ecosystem may inhibit algae growth and subsequently lead to the decline of primary productivity. Growth inhibition tests are required for ecotoxicological assessments for regulatory purposes. In silico study is playing an important role in replacing or reducing animal tests and decreasing experimental expense due to its efficiency. In this work, a series of theoretical models was developed for predicting algal growth inhibition (log EC₅₀) after 72 h exposure to diverse chemicals. In total 348 organic compounds were classified into five modes of toxic action using the Verhaar Scheme. Each model was established by using molecular descriptors that characterize electronic and structural properties. The external validation and leave-one-out cross validation proved the statistical robustness of the derived models. Thus they can be used to predict log EC₅₀ values of chemicals that lack authorized algal growth inhibition values (72 h). This work systematically studied algal growth inhibition according to toxic modes and the developed model suite covers all five toxic modes. The outcome of this research will promote toxic mechanism analysis and be made applicable to structural diversity.

In silico prediction of pesticide aquatic toxicity with chemical category approaches

Aquatic toxicity is an important issue in pesticide development. In this study, using nine molecular fingerprints to describe pesticides, binary and ternary classification models were constructed to predict aquatic toxicity of pesticides via six machine learning methods: Naïve Bayes (NB), Artificial Neural Network (ANN), k-Nearest Neighbor (kNN), Classification Tree (CT), Random Forest (RF) and Support Vector Machine (SVM). For the binary models, local models were obtained with 829 pesticides on rainbow trout (RT) and 151 pesticides on lepomis (LP), and global models were constructed on the basis of 1258 diverse pesticides on RT and LP and 278 on other fish species. After analyzing the local binary models, we found that fish species caused influence in terms of accuracy. Considering the data size and predictive range, the 1258 pesticides were also used to build global ternary models. The best local binary models were Maccs_ANN for RT and Maccs_SVM for LP, which exhibited accuracies of 0.90 and 0.90, respectively. For global binary models, the best model was Graph_SVM with an accuracy of 0.89. Accuracy of the best global ternary model Graph_SVM was 0.81, which was a little lower than that of the best global binary model. In addition, several substructural alerts were identified including nitrobenzene, chloroalkene and nitrile, which could significantly correlate with pesticide aquatic toxicity. This study provides a useful tool for an early evaluation of pesticide aquatic toxicity in environmental risk assessment.

Mode of Action (MOA) Assignment Classifications for Ecotoxicology: An Evaluation of Approaches

The mode of toxic action (MOA) is recognized as a key determinant of chemical toxicity and as an alternative to chemical class-based predictive toxicity modeling. However, MOA classification has never been standardized in ecotoxicology, and a comprehensive comparison of classification tools and approaches has never been reported. Here we critically evaluate three MOA classification methodologies using an aquatic toxicity data set of 3448 chemicals, compare the approaches, and assess utility and limitations in screening and early tier assessments. The comparisons focused on three commonly used tools: Verhaar prediction of toxicity MOA, the U.S. Environmental Protection Agency (EPA) ASsessment Tool for Evaluating Risk (ASTER) QSAR (quantitative structure activity relationship) application, and the EPA Mode of Action and Toxicity (MOAtox) database. Of the 3448 MOAs predicted using the Verhaar scheme, 1165 were classified by ASTER, and 802 were available in MOAtox. Of the subset of 432 chemicals with MOA assignments for each of the three schemes, 42% had complete concordance in MOA classification, and there was no agreement for 7% of the chemicals. The research shows the potential for large differences in MOA classification between the five broad groups of the Verhaar scheme and the more mechanism-based assignments of ASTER and MOAtox. Harmonization of classification schemes is needed to use MOA classification in chemical hazard and risk assessment more broadly.

Prediction of pesticide acute toxicity using two-dimensional chemical descriptors and target species classification

Previous modelling of the median lethal dose (oral rat LD₅₀) has indicated that local class-based models yield better correlations than global models. We evaluated the hypothesis that dividing the dataset by pesticidal mechanisms would improve prediction accuracy. A linear discriminant analysis (LDA) based-approach was utilized to assign indicators such as the pesticide target species, mode of action, or target species - mode of action combination. LDA models were able to predict these indicators with about 87% accuracy. Toxicity is predicted utilizing the QSAR model fit to chemicals with that indicator. Toxicity was also predicted using a global hierarchical clustering (HC) approach which divides data set into clusters based on molecular similarity. At a comparable prediction coverage (~94%), the global HC method yielded slightly higher prediction accuracy (r² = 0.50) than the LDA method (r² ~ 0.47). A single model fit to the entire training set yielded the poorest results (r² = 0.38), indicating that there is an advantage to clustering the dataset to predict acute toxicity. Finally, this study shows that whilst dividing the training set into subsets (i.e. clusters) improves prediction accuracy, it may not matter which method (expert based or purely machine learning) is used to divide the dataset into subsets.

A History of the Molecular Initiating Event

The adverse outcome pathway (AOP) framework provides an alternative to traditional in vivo experiments for the risk assessment of chemicals. AOPs consist of a number of key events (KEs) linked by key event relationships across a range of biological organization backed by scientific evidence. The first KE in the pathway is the molecular initiating event (MIE)-the initial chemical trigger that starts an AOP. Over the past 3 years the AOP conceptual framework has gained a large amount of momentum in toxicology as an alternative to animal methods, and so the MIE has come into the spotlight. What is an MIE? How can MIEs be measured or predicted? What research is currently contributing to our understanding of MIEs? In this Perspective we outline answers to these key questions.

A Bayesian network model for predicting aquatic toxicity mode of action using two dimensional theoretical molecular descriptors

The mode of toxic action (MoA) has been recognized as a key determinant of chemical toxicity, but development of predictive MoA classification models in aquatic toxicology has been limited. We developed a Bayesian network model to classify aquatic toxicity MoA using a recently published dataset containing over one thousand chemicals with MoA assignments for aquatic animal toxicity. Two dimensional theoretical chemical descriptors were generated for each chemical using the Toxicity Estimation Software Tool. The model was developed through augmented Markov blanket discovery from the dataset of 1098 chemicals with the MoA broad classifications as a target node. From cross validation, the overall precision for the model was 80.2%. The best precision was for the AChEI MoA (93.5%) where 257 chemicals out of 275 were correctly classified. Model precision was poorest for the reactivity MoA (48.5%) where 48 out of 99 reactive chemicals were correctly classified. Narcosis represented the largest class within the MoA dataset and had a precision and reliability of 80.0%, reflecting the global precision across all of the MoAs. False negatives for narcosis most often fell into electron transport inhibition, neurotoxicity or reactivity MoAs. False negatives for all other MoAs were most often narcosis. A probabilistic sensitivity analysis was undertaken for each MoA to examine the sensitivity to individual and multiple descriptor findings. The results show that the Markov blanket of a structurally complex dataset can simplify analysis and interpretation by identifying a subset of the key chemical descriptors associated with broad aquatic toxicity MoAs, and by providing a computational chemistry-based network classification model with reasonable prediction accuracy.

Predicting the aquatic toxicity mode of action using logistic regression and linear discriminant analysis

The paper highlights the use of the logistic regression (LR) method in the construction of acceptable statistically significant, robust and predictive models for the classification of chemicals according to their aquatic toxic modes of action. Essentials accounting for a reliable model were all considered carefully. The model predictors were selected by stepwise forward discriminant analysis (LDA) from a combined pool of experimental data and chemical structure-based descriptors calculated by the CODESSA and DRAGON software packages. Model predictive ability was validated both internally and externally. The applicability domain was checked by the leverage approach to verify prediction reliability. The obtained models are simple and easy to interpret. In general, LR performs much better than LDA and seems to be more attractive for the prediction of the more toxic compounds, i.e. compounds that exhibit excess toxicity versus non-polar narcotic compounds and more reactive compounds versus less reactive compounds. In addition, model fit and regression diagnostics was done through the influence plot which reflects the hat-values, studentized residuals, and Cook's distance statistics of each sample. Overdispersion was also checked for the LR model. The relationships between the descriptors and the aquatic toxic behaviour of compounds are also discussed.

Prediction of in vitro and in vivo oestrogen receptor activity using hierarchical clustering

In this study, hierarchical clustering classification models were developed to predict in vitro and in vivo oestrogen receptor (ER) activity. Classification models were developed for binding, agonist, and antagonist in vitro ER activity and for mouse in vivo uterotrophic ER binding. In vitro classification models yielded balanced accuracies ranging from 0.65 to 0.85 for the external prediction set. In vivo ER classification models yielded balanced accuracies ranging from 0.72 to 0.83. If used as additional biological descriptors for in vivo models, in vitro scores were found to increase the prediction accuracy of in vivo ER models. If in vitro activity was used directly as a surrogate for in vivo activity, the results were poor (balanced accuracy ranged from 0.49 to 0.72). Under-sampling negative compounds in the training set was found to increase the coverage (fraction of chemicals which can be predicted) and increase prediction sensitivity.

Modeling the toxicity of chemical pesticides in multiple test species using local and global QSTR approaches

The safety assessment processes require the toxicity data of chemicals in multiple test species and thus, emphasize the need for computational methods capable of toxicity prediction in multiple test species. Pesticides are designed toxic substances and find extensive applications worldwide. In this study, we have established local and global QSTR (quantitative structure-toxicity relationship) and ISC QSAAR (interspecies correlation quantitative structure activity-activity relationship) models for predicting the toxicities of pesticides in multiple aquatic test species using the toxicity data in crustacean (Daphnia magna, Americamysis bahia, Gammarus fasciatus, and Penaeus duorarum) and fish (Oncorhynchus mykiss and Lepomis macrochirus) species in accordance with the OECD guidelines. The ensemble learning based QSTR models (decision tree forest, DTF and decision tree boost, DTB) were constructed and validated using several statistical coefficients derived on the test data. In all the QSTR and QSAAR models, Log P was an important predictor. The constructed local, global and interspecies QSAAR models yielded high correlations (R2) of >0.941; >0.943 and >0.826, respectively between the measured and model predicted endpoint toxicity values in the test data. The performances of the local and global QSTR models were comparable. Furthermore, the chemical applicability domains of these QSTR/QSAAR models were determined using the leverage and standardization approaches. The results suggest for the appropriateness of the developed QSTR/QSAAR models to reliably predict the aquatic toxicity of structurally diverse pesticides in multiple test species and can be used for the screening and prioritization of new pesticides.

MOAtox: A comprehensive mode of action and acute aquatic toxicity database for predictive model development

The mode of toxic action (MOA) has been recognized as a key determinant of chemical toxicity and as an alternative to chemical class-based predictive toxicity modeling. However, the development of quantitative structure activity relationship (QSAR) and other models has been limited by the availability of comprehensive high quality MOA and toxicity databases. The current study developed a dataset of MOA assignments for 1213 chemicals that included a diversity of metals, pesticides, and other organic compounds that encompassed six broad and 31 specific MOAs. MOA assignments were made using a combination of high confidence approaches that included international consensus classifications, QSAR predictions, and weight of evidence professional judgment based on an assessment of structure and literature information. A toxicity database of 674 acute values linked to chemical MOA was developed for fish and invertebrates. Additionally, species-specific measured or high confidence estimated acute values were developed for the four aquatic species with the most reported toxicity values: rainbow trout (Oncorhynchus mykiss), fathead minnow (Pimephales promelas), bluegill (Lepomis macrochirus), and the cladoceran (Daphnia magna). Measured acute toxicity values met strict standardization and quality assurance requirements. Toxicity values for chemicals with missing species-specific data were estimated using established interspecies correlation models and procedures (Web-ICE; http://epa.gov/ceampubl/fchain/webice/), with the highest confidence values selected. The resulting dataset of MOA assignments and paired toxicity values are provided in spreadsheet format as a comprehensive standardized dataset available for predictive aquatic toxicology model development.

Comparison of global and mode of action-based models for aquatic toxicity

The ability to estimate aquatic toxicity is a critical need for ecological risk assessment and chemical regulation. The consensus in the literature is that mode of action (MOA) based toxicity models yield the most toxicologically meaningful and, theoretically, the most accurate results. In this study, a two-step prediction methodology was developed to estimate acute aquatic toxicity from molecular structure. In the first step, one-against-the-rest linear discriminant analysis (LDA) models were used to predict the MOA. The LDA models were able to predict the MOA with 85.8-88.8% accuracy for broad and specific MOAs, respectively. In the second step, a multiple linear regression (MLR) model corresponding to the predicted MOA was used to predict the acute aquatic toxicity value. The MOA-based approach was found to yield similar external prediction accuracy (r(2) = 0.529-0.632) to a single global MLR model (r(2) = 0.551-0.562) fit to the entire training set. Overall, the global hierarchical clustering approach yielded a higher combination of accuracy and prediction coverage (r(2) = 0.572, coverage = 99.3%) than the other approaches. Utilizing multiple two-dimensional chemical descriptors in MLR models yielded comparable results to using only the octanol-water partition coefficient (log K(ow)).

In silico prediction of chemical aquatic toxicity with chemical category approaches and substructural alerts

Aquatic toxicity is an important endpoint in the evaluation of chemically adverse effects on ecosystems.