QSAR models for reproductive toxicity and endocrine disruption activity

Interpretable machine learning for the identification of estrogen receptor agonists, antagonists, and binders

An abnormal hormonal activity or exposure to endocrine-disrupting chemicals (EDCs) can cause endocrine system malfunction. Among the many interactions EDCs can affect is the disruption of estrogen signalling, which can lead to adverse health effects such as cancer, osteoporosis, neurodegenerative diseases, cardiovascular disease, insulin resistance, and obesity. Knowing which chemical can act as an EDC is a significant advantage and a practical necessity. New Approach Methodologies (NAM) computational models offer a quick and cost-effective solution for preliminary hazard assessment of chemicals without animal testing. Therefore, a machine learning approach was used to investigate the relationships between estrogen receptor (ER) activity and chemical structure to identify chemicals that can interact with ER. For this purpose, the consolidated in vitro assay data from ToxCast/Tox21 projects was used for developing Random Forest classification models for ER binding, agonists, and antagonists. The overall classification prediction accuracy reaches up to 82%, depending on whether the model predicted agonists, antagonists, or compounds that bind to the active site. Given the imbalance in endocrine disruption data, the derived models are good candidates for deprioritising chemicals and reducing animal testing. The interpretation of theoretical molecular descriptors of the models was consistent with the molecular interactions known in the ligand binding pocket. The estimated class probabilities enabled the analysis of the applicability domain of the developed models and the assessment of the predictions' reliability, followed by the guidelines for interpreting prediction results. The models are openly accessible and useable at QsarDB.org (http://dx.doi.org/10.15152/QDB.259) according to the FAIR (Findable, Accessible, Interoperable, Reusable) principles.

Machine learning models for rat multigeneration reproductive toxicity prediction

Reproductive toxicity is one of the prominent endpoints in the risk assessment of environmental and industrial chemicals. Due to the complexity of the reproductive system, traditional reproductive toxicity testing in animals, especially guideline multigeneration reproductive toxicity studies, take a long time and are expensive. Therefore, machine learning, as a promising alternative approach, should be considered when evaluating the reproductive toxicity of chemicals. We curated rat multigeneration reproductive toxicity testing data of 275 chemicals from ToxRefDB (Toxicity Reference Database) and developed predictive models using seven machine learning algorithms (decision tree, decision forest, random forest, k-nearest neighbors, support vector machine, linear discriminant analysis, and logistic regression). A consensus model was built based on the seven individual models. An external validation set was curated from the COSMOS database and the literature. The performances of individual and consensus models were evaluated using 500 iterations of 5-fold cross-validations and the external validation data set. The balanced accuracy of the models ranged from 58% to 65% in the 5-fold cross-validations and 45%–61% in the external validations. Prediction confidence analysis was conducted to provide additional information for more appropriate applications of the developed models. The impact of our findings is in increasing confidence in machine learning models. We demonstrate the importance of using consensus models for harnessing the benefits of multiple machine learning models (i.e., using redundant systems to check validity of outcomes). While we continue to build upon the models to better characterize weak toxicants, there is current utility in saving resources by being able to screen out strong reproductive toxicants before investing in vivo testing. The modeling approach (machine learning models) is offered for assessing the rat multigeneration reproductive toxicity of chemicals. Our results suggest that machine learning may be a promising alternative approach to evaluate the potential reproductive toxicity of chemicals.

Ensemble QSAR Modeling to Predict Multispecies Fish Toxicity Lethal Concentrations and Points of Departure

QSAR modeling can be used to aid testing prioritization of the thousands of chemical substances for which no ecological toxicity data are available. We drew on the U.S. Environmental Protection Agency's ECOTOX database with additional data from ECHA to build a large data set containing in vivo test data on fish for thousands of chemical substances. This was used to create QSAR models to predict two types of end points: acute LC50 (median lethal concentration) and points of departure similar to the NOEC (no observed effect concentration) for any duration (named the "LC50" and "NOEC" models, respectively). These models used study covariates, such as species and exposure route, as features to facilitate the simultaneous use of varied data types. A novel method of substituting taxonomy groups for species dummy variables was introduced to maximize generalizability to different species. A stacked ensemble of three machine learning methods-random forest, gradient boosted trees, and support vector regression-was implemented to best make use of a large data set with many descriptors. The LC50 and NOEC models predicted end points within 1 order of magnitude 81% and 76% of the time, respectively, and had RMSEs of roughly 0.83 and 0.98 log10(mg/L), respectively. Benchmarks against the existing TEST and ECOSAR tools suggest improved prediction accuracy.

Molecular Descriptors for Structure-Activity Applications: A Hands-On Approach

Molecular descriptors capture diverse parts of the structural information of molecules and they are the support of many contemporary computer-assisted toxicological and chemical applications. After briefly introducing some fundamental concepts of structure-activity applications (e.g., molecular descriptor dimensionality, classical vs. fingerprint description, and activity landscapes), this chapter guides the readers through a step-by-step explanation of molecular descriptors rationale and application. To this end, the chapter illustrates a case study of a recently published application of molecular descriptors for modeling the activity on cytochrome P450.

Integration of in silico methods and computational systems biology to explore endocrine-disrupting chemical binding with nuclear hormone receptors

Thousands of potential endocrine-disrupting chemicals present difficult regulatory challenges. Endocrine-disrupting chemicals can interfere with several nuclear hormone receptors associated with a variety of adverse health effects. The U.S. Environmental Protection Agency (U.S. EPA) has released its reviews of Tier 1 screening assay results for a set of pesticides in the Endocrine Disruptor Screening Program (EDSP), and recently, the Collaborative Estrogen Receptor Activity Prediction Project (CERAPP) data. In this study, the predictive ability of QSAR and docking approaches is evaluated using these data sets. This study also presents a computational systems biology approach using carbaryl (1-naphthyl methylcarbamate) as a case study. For estrogen receptor and androgen receptor binding predictions, two commercial and two open source QSAR tools were used, as was the publicly available docking tool Endocrine Disruptome. For estrogen receptor binding predictions, the ADMET Predictor, VEGA, and OCHEM models (specificity: 0.88, 0.88, and 0.86, and accuracy: 0.81, 0.84, and 0.88, respectively) were each more reliable than the MetaDrug™ model (specificity 0.81 and accuracy 0.77). For androgen receptor binding predictions, the Endocrine Disruptome and ADMET Predictor models (specificity: 0.94 and 0.8, and accuracy: 0.78 and 0.71, respectively) were more reliable than the MetaDrug™ model (specificity 0.33 and accuracy 0.4). A consensus approach is proposed that reaches general agreement among the models (specificity 0.94 and accuracy 0.89). This study integrates QSAR, docking, and systems biology approaches as a virtual screening tool for use in risk assessment. As such, this systems biology pathways and network analysis approach provides a means to more critically assess the potential effects of endocrine-disrupting chemicals.

In Silico Prediction for Intestinal Absorption and Brain Penetration of Chemical Pesticides in Humans

Intestinal absorption and brain permeation constitute key parameters of toxicokinetics for pesticides, conditioning their toxicity, including neurotoxicity. However, they remain poorly characterized in humans. The present study was therefore designed to evaluate human intestine and brain permeation for a large set of pesticides (n = 338) belonging to various chemical classes, using an in silico graphical BOILED-Egg/SwissADME online method based on lipophilicity and polarity that was initially developed for drugs. A high percentage of the pesticides (81.4%) was predicted to exhibit high intestinal absorption, with a high accuracy (96%), whereas a lower, but substantial, percentage (38.5%) displayed brain permeation. Among the pesticide classes, organochlorines (n = 30) constitute the class with the lowest percentage of intestine-permeant members (40%), whereas that of the organophosphorus compounds (n = 99) has the lowest percentage of brain-permeant chemicals (9%). The predictions of the permeations for the pesticides were additionally shown to be significantly associated with various molecular descriptors well-known to discriminate between permeant and non-permeant drugs. Overall, our in silico data suggest that human exposure to pesticides through the oral way is likely to result in an intake of these dietary contaminants for most of them and brain permeation for some of them, thus supporting the idea that they have toxic effects on human health, including neurotoxic effects.

Rigorous 3-dimensional spectral data activity relationship approach modeling strategy for ToxCast estrogen receptor data classification, validation, and feature extraction

The estrogenic potential (expressed as a score composite of 18 high throughput screening bioassays) of 1528 compounds from the ToxCast database was modeled by a 3-dimensional spectral data activity relationship approach (3D-SDAR). Due to a lack of ¹⁷ O nuclear magnetic resonance (NMR) simulation software, the most informative carbon-carbon 3D-SDAR fingerprints were augmented with indicator variables representing oxygen atoms from carbonyl and carboxamide, ester, sulfonyl, nitro, aliphatic hydroxyl, and phenolic hydroxyl groups. To evaluate the true predictive performance of the authors' model the United States Environmental Protection Agency provided them with a blind test set consisting of 2008 compounds. Of these, 543 had available literature data-their binding affinity served to estimate the external classification accuracy of the developed model: predictive accuracy of 0.62, sensitivity of 0.71, and specificity of 0.53 were obtained. Compared with alternative modeling techniques, the authors' model displayed very little reduction in performance between the modeling and the prediction set. A 3D-SDAR mapping technique allowed identification of structural features essential for estrogenicity: 1) the presence of a phenolic OH group or cyclohexenone, 2) a second aromatic or phenolic ring at a distance of 6 Å to 8 Å from the oxygen of the first phenol ring, 3) the presence of a methyl group approximately 6 Å away from the centroid of a phenol ring, and 4) a carbonyl group in close proximity (∼4 Å measured to the centroid) to 1 of the phenol rings. Environ Toxicol Chem 2017;36:823-830. Published 2016 Wiley Periodicals Inc. on behalf of SETAC. This article is a US government work and, as such, is in the public domain in the United States of America.

The connection Between Plasma Protein Binding and Acute Toxicity as Determined by the LD50 Value

Preclinical Research A dataset of three drug classes (acids, bases, and neutrals) with LD50 values in mice was analysed to investigate a possible connection between high plasma protein binding and acute toxicity. Initially, it was found that high plasma protein binding was associated with toxicity for acids and neutrals, but after compensating for differences in lipophilicity, plasma protein binding was found not to be associated with toxicity. The therapeutic index established by the quotient between mouse LD50 and the defined daily dose was unaffected by both lipophilicity and plasma protein binding.

Computational study of binding affinity to nuclear receptors for some cosmetic ingredients

We studied the ingredients of cosmetic products as potential endocrine disruptors (ED) by in silico methods (docking). The structures of 14 human nuclear receptors have been retrieved from the protein data bank (PDB). We only considered the mechanism linked with direct binding to nuclear receptors with well-defined crystal structures. Predictions were performed using the Endocrine Disruptome docking program http://endocrinedisruptome.ki.si/ (Kolšek et al., 2013). 122 compounds were estimated to be possible endocrine disruptors bind to at least one of the receptors, 21 of them which are predicted to be probable toxicants for endocrine disruption as they bind to more than five receptors simultaneously. According to the literature survey and lack of experimental data it remains a challenge to prove or disprove the in silico results experimentally also for other potential endocrine disruptors.

The ChemScreen project to design a pragmatic alternative approach to predict reproductive toxicity of chemicals

There is a great need for rapid testing strategies for reproductive toxicity testing, avoiding animal use. The EU Framework program 7 project ChemScreen aimed to fill this gap in a pragmatic manner preferably using validated existing tools and place them in an innovative alternative testing strategy. In our approach we combined knowledge on critical processes affected by reproductive toxicants with knowledge on the mechanistic basis of such effects. We used in silico methods for prescreening chemicals for relevant toxic effects aiming at reduced testing needs. For those chemicals that need testing we have set up an in vitro screening panel that includes mechanistic high throughput methods and lower throughput assays that measure more integrative endpoints. In silico pharmacokinetic modules were developed for rapid exposure predictions via diverse exposure routes. These modules to match in vitro and in vivo exposure levels greatly improved predictivity of the in vitro tests. As a further step, we have generated examples how to predict reproductive toxicity of chemicals using available data. We have executed formal validations of panel constituents and also used more innovative manners to validate the test panel using mechanistic approaches. We are actively engaged in promoting regulatory acceptance of the tools developed as an essential step towards practical application, including case studies for read-across purposes. With this approach, a significant saving in animal use and associated costs seems very feasible.

Applying Adverse Outcome Pathways (AOPs) to support Integrated Approaches to Testing and Assessment (IATA)

Chemical regulation is challenged by the large number of chemicals requiring assessment for potential human health and environmental impacts. Current approaches are too resource intensive in terms of time, money and animal use to evaluate all chemicals under development or already on the market. The need for timely and robust decision making demands that regulatory toxicity testing becomes more cost-effective and efficient. One way to realize this goal is by being more strategic in directing testing resources; focusing on chemicals of highest concern, limiting testing to the most probable hazards, or targeting the most vulnerable species. Hypothesis driven Integrated Approaches to Testing and Assessment (IATA) have been proposed as practical solutions to such strategic testing. In parallel, the development of the Adverse Outcome Pathway (AOP) framework, which provides information on the causal links between a molecular initiating event (MIE), intermediate key events (KEs) and an adverse outcome (AO) of regulatory concern, offers the biological context to facilitate development of IATA for regulatory decision making. This manuscript summarizes discussions at the Workshop entitled "Advancing AOPs for Integrated Toxicology and Regulatory Applications" with particular focus on the role AOPs play in informing the development of IATA for different regulatory purposes.

A novel approach to generate robust classification models to predict developmental toxicity from imbalanced datasets

Computational models to predict the developmental toxicity of compounds are built on imbalanced datasets wherein the toxicants outnumber the non-toxicants. Consequently, the results are biased towards the majority class (toxicants). To overcome this problem and to obtain sensitive but also accurate classifiers, we followed an integrated approach wherein (i) Synthetic Minority Over Sampling (SMOTE) is used for re-sampling, (ii) genetic algorithm (GA) is used for variable selection and (iii) support vector machines (SVM) is used for model development. The best model, M3, has (i) sensitivity (SE) = 85.54% and specificity (SP) = 85.62% in leave-one-out validation, (ii) classification accuracy of the training set = 99.67%, (iii) classification accuracy of the test set = 92.59%; and (iv) sensitivity = 92.68, specificity = 92.31 on the test set. Consensus prediction based on models M3-M5 improved these percentages by 5% over M3. From the analysis of results we infer that data imbalance in toxicity studies can be effectively addressed by the application of re-sampling techniques.

Classification models for safe drug molecules

Frequent failure of drug candidates during development stages remains the major deterrent for an early introduction of new drug molecules. The drug toxicity is the major cause of expensive late-stage development failures. An early identification/optimization of the most favorable molecule will naturally save considerable cost, time, human efforts and minimize animal sacrifice. (Quantitative) Structure Activity Relationships [(Q)SARs] represent statistically derived predictive models correlating biological activity (including desirable therapeutic effect and undesirable side effects) of chemicals (drugs/toxicants/environmental pollutants) with molecular descriptors and/or properties. (Q)SAR models which categorize the available data into two or more groups/classes are known as classification models. Numerous techniques of diverse nature are being presently employed for development of classification models. Though there is an increasing use of classification models for prediction of either biological activity or toxicity, the future trend will naturally be towards the development of classification models capable of simultaneous prediction of biological activity, toxicity, and pharmacokinetic parameters so as to accelerate development of bioavailable safe drug molecules.

Application of an integrated testing strategy to the U.S. EPA endocrine disruptor screening program

New approaches to generating and evaluating toxicity data for chemicals are needed to cope with the ever-increasing demands of new programs. One such approach involves the use of an integrated testing and evaluation strategy based on the specific properties and activities of a chemical. Such an integrated strategy, whether applied to existing or future programs, can promote efficient use of resources and save animals. We demonstrate the utility of such a strategy by applying it to the current U.S. Environmental Protection Agency Endocrine Disruptor Screening Program (EDSP). Launched in October 2009, the EDSP utilizes a two-tiered approach, whereby each tier requires a battery of animal-intensive and expensive tests. Tier 1 consists of five in vitro and six in vivo assays that are intended to determine a chemical's potential to interact with the estrogen (E), androgen (A), or thyroid (T) hormone pathways. Tier 2 is proposed to consist of multigenerational reproductive and developmental toxicity tests in several species and is intended to determine whether a chemical can cause adverse effects resulting from E, A, or T modulation. In contrast to the existing EDSP structure, we show, using the pesticide atrazine as an example, that a multilevel testing framework combined with an integrated evaluation process would significantly increase efficiency by minimizing testing.

An integrated QSAR-PBPK modelling approach for predicting the inhalation toxicokinetics of mixtures of volatile organic chemicals in the rat

The objective of this study was to predict the inhalation toxicokinetics of chemicals in mixtures using an integrated QSAR-PBPK modelling approach. The approach involved: (1) the determination of partition coefficients as well as V(max) and K(m) based solely on chemical structure for 53 volatile organic compounds, according to the group contribution approach; and (2) using the QSAR-driven coefficients as input in interaction-based PBPK models in the rat to predict the pharmacokinetics of chemicals in mixtures of up to 10 components (benzene, toluene, m-xylene, o-xylene, p-xylene, ethylbenzene, dichloromethane, trichloroethylene, tetrachloroethylene, and styrene). QSAR-estimated values of V(max) varied compared with experimental results by a factor of three for 43 out of 53 studied volatile organic compounds (VOCs). K(m) values were within a factor of three compared with experimental values for 43 out of 53 VOCs. Cross-validation performed as a ratio of predicted residual sum of squares and sum of squares of the response value indicates a value of 0.108 for V(max) and 0.208 for K(m). The integration of QSARs for partition coefficients, V(max) and K(m), as well as setting the K(m) equal to K(i) (metabolic inhibition constant) within the mixture PBPK model allowed to generate simulations of the inhalation pharmacokinetics of benzene, toluene, m-xylene, o-xylene, p-xylene, ethylbenzene, dichloromethane, trichloroethylene, tetrachloroethylene and styrene in various mixtures. Overall, the present study indicates the potential usefulness of the QSAR-PBPK modelling approach to provide first-cut evaluations of the kinetics of chemicals in mixtures of increasing complexity, on the basis of chemical structure.

Computational estimation of rainbow trout estrogen receptor binding affinities for environmental estrogens

Environmental estrogens have been the subject of intense research due to their documented detrimental effects on the health of fish and wildlife and their potential to negatively impact humans. A complete understanding of how these compounds affect health is complicated because environmental estrogens are a structurally heterogeneous group of compounds. In this work, computational molecular dynamics simulations were utilized to predict the binding affinity of different compounds using rainbow trout (Oncorhynchus mykiss) estrogen receptors (ERs) as a model. Specifically, this study presents a comparison of the binding affinity of the natural ligand estradiol-17β to the four rainbow trout ER isoforms with that of three known environmental estrogens 17α-ethinylestradiol, bisphenol A, and raloxifene. Two additional compounds, atrazine and testosterone, that are known to be very weak or non-binders to ERs were tested. The binding affinity of these compounds to the human ERα subtype is also included for comparison. The results of this study suggest that, when compared to estradiol-17β, bisphenol A binds less strongly to all four receptors, 17α-ethinylestradiol binds more strongly, and raloxifene has a high affinity for the α subtype only. The results also show that atrazine and testosterone are weak or non-binders to the ERs. All of the results are in excellent qualitative agreement with the known in vivo estrogenicity of these compounds in the rainbow trout and other fishes. Computational estimation of binding affinities could be a valuable tool for predicting the impact of environmental estrogens in fish and other animals.

Prediction of toxicity and data exploratory analysis of estrogen-active endocrine disruptors using counter-propagation artificial neural networks

In this work, a novel algorithm for optimization of counter-propagation artificial neural networks has been used for development of quantitative structure-activity relationships model for prediction of the estrogenic activity of endocrine-disrupting chemicals. The search for the best model was performed using genetic algorithms. Genetic algorithms were used not only for selection of the most suitable descriptors for modeling, but also for automatic adjustment of their relative importance. Using our recently developed algorithm for automatic adjustment of the relative importance of the input variables, we have developed simple models with very good generalization performances using only few interpretable descriptors. One of the developed models is in details discussed in this article. The simplicity of the chosen descriptors and their relative importance for this model helped us in performing a detailed data exploratory analysis which gave us an insight in the structural features required for the activity of the estrogenic endocrine-disrupting chemicals.