A new hybrid system of QSAR models for predicting bioconcentration factors (BCF)

Integrated Transfer Learning and Multitask Learning Strategies to Construct Graph Neural Network Models for Predicting Bioaccumulation Parameters of Chemicals

Accurate prediction of parameters related to the environmental exposure of chemicals is crucial for the sound management of chemicals. However, the lack of large data sets for training models may result in poor prediction accuracy and robustness. Herein, integrated transfer learning (TL) and multitask learning (MTL) was proposed for constructing a graph neural network (GNN) model (abbreviated as TL-MTL-GNN model) using n-octanol/water partition coefficients as a source domain. The TL-MTL-GNN model was trained to predict three bioaccumulation parameters based on enlarged data sets that cover 2496 compounds with at least one bioaccumulation parameter. Results show that the TL-MTL-GNN model outperformed single-task GNN models with and without the TL, as well as conventional machine learning models trained with molecular descriptors or fingerprints. Applicability domains were characterized by a state-of-the-art structure-activity landscape-based (abbreviated as ADSAL) methodology. The TL-MTL-GNN model coupled with the optimal ADSAL was employed to predict bioaccumulation parameters for around 60,000 chemicals, with more than 13,000 compounds identified as bioaccumulative chemicals. The high predictive accuracy and robustness of the TL-MTL-GNN model demonstrate the feasibility of integrating the TL and MTL strategy in modeling small-sized data sets. The strategy holds significant potential for addressing small data challenges in modeling environmental chemicals.

ChemFREE: a one-stop comprehensive platform for ecological and environmental risk evaluation of chemicals in one health world

Addressing health and safety crises stemming from various environmental and ecological issues is a core focus of One Health (OH), which aims to balance and optimize the health of humans, animals, and the environment. While many chemicals contribute significantly to our quality of life when properly used, others pose environmental and ecological health risks. Recently, assessing the ecological and environmental risks associated with chemicals has gained increasing significance in the OH world. In silico models may address time-consuming and costly challenges, and fill gaps in situations where no experimental data is available. However, despite their significant contributions, these assessment models are not web-integrated, leading to user inconvenience. In this study, we developed a one-stop comprehensive web platform for freely evaluating the eco-environmental risk of chemicals, named ChemFREE (Chemical Formula Risk Evaluation of Eco-environment, available in http://chemfree.agroda.cn/chemfree/). Inputting SMILES string of chemicals, users will obtain the assessment outputs of ecological and environmental risk, etc. A performance evaluation of 2935 external chemicals revealed that most classification models achieved an accuracy rate above 0.816. Additionally, the $Q_{F1}^2$ metric for regression models ranges from 0.618 to 0.898. Therefore, it will facilitate the eco-environmental risk evaluation of chemicals in the OH world.

Ecotoxicological Evaluation of Bisphenol A and Alternatives: A Comprehensive In Silico Modelling Approach

Bisphenol A (BPA), a compound widely used in industrial applications, has raised concerns due to its environmental impact. As a key component in the manufacture of polycarbonate plastics and epoxy resins used in many consumer products, concerns about potential harm to human health and the environment are unavoidable. This study seeks to address these concerns by evaluating a range of potential BPA alternatives, focusing on their ecotoxicological properties. The research examines 76 bisphenols, including BPA derivatives, using a variety of in silico ecotoxicological models, although it should be noted that these models were not developed exclusively for this particular class of compounds. Consequently, interpretations should be made with caution. The results of this study highlight specific compounds of potential environmental concern and underscore the need to develop more specific models for BPA alternatives that will allow for more accurate and reliable assessment.

QSAR Models for the Prediction of Dietary Biomagnification Factor in Fish

Xenobiotics released in the environment can be taken up by aquatic and terrestrial organisms and can accumulate at higher concentrations through the trophic chain. Bioaccumulation is therefore one of the PBT properties that authorities require to assess for the evaluation of the risks that chemicals may pose to humans and the environment. The use of an integrated testing strategy (ITS) and the use of multiple sources of information are strongly encouraged by authorities in order to maximize the information available and reduce testing costs. Moreover, considering the increasing demand for development and the application of new approaches and alternatives to animal testing, the development of in silico cost-effective tools such as QSAR models becomes increasingly important. In this study, a large and curated literature database of fish laboratory-based values of dietary biomagnification factor (BMF) was used to create externally validated QSARs. The quality categories (high, medium, low) available in the database were used to extract reliable data to train and validate the models, and to further address the uncertainty in low-quality data. This procedure was useful for highlighting problematic compounds for which additional experimental effort would be required, such as siloxanes, highly brominated and chlorinated compounds. Two models were suggested as final outputs in this study, one based on good-quality data and the other developed on a larger dataset of consistent Log BMF_L values, which included lower-quality data. The models had similar predictive ability; however, the second model had a larger applicability domain. These QSARs were based on simple MLR equations that could easily be applied for the predictions of dietary BMF_L in fish, and support bioaccumulation assessment procedures at the regulatory level. To ease the application and dissemination of these QSARs, they were included with technical documentation (as QMRF Reports) in the QSAR-ME Profiler software for QSAR predictions available online.

Affinity of Compounds for Phosphatydylcholine-Based Immobilized Artificial Membrane-A Measure of Their Bioconcentration in Aquatic Organisms

The BCF (bioconcentration factor) of solutes in aquatic organisms is an important parameter because many undesired chemicals enter the ecosystem and affect the wildlife. Chromatographic retention factor log k_w^IAM obtained from immobilized artificial membrane (IAM) HPLC chromatography with buffered, aqueous mobile phases and calculated molecular descriptors obtained for a group of 120 structurally unrelated compounds were used to generate useful models of log BCF. It was established that log k_w^IAM obtained in the conditions described in this study is not sufficient as a sole predictor of bioconcentration. Simple, potentially useful models based on log k_w^IAM and a selection of readily available, calculated descriptors and accounting for over 88% of total variability were generated using multiple linear regression (MLR), partial least squares (PLS) regression and artificial neural networks (ANN). The models proposed in the study were tested on an external group of 120 compounds and on a group of 40 compounds with known experimental log BCF values. It was established that a relatively simple MLR model containing four independent variables leads to satisfying BCF predictions and is more intuitive than PLS or ANN models.

Predicting the Bioconcentration Factor in Fish from Molecular Structures

The bioconcentration factor (BCF) is one of the metrics used to evaluate the potential of a substance to bioaccumulate into aquatic organisms. In this work, linear and non-linear regression QSARs were developed for the prediction of log BCF using different computational approaches, and starting from a large and structurally heterogeneous dataset. The new MLR-OLS and ANN regression models have good fitting with R2 values of 0.62 and 0.70, respectively, and comparable external predictivity with R2ext 0.64 and 0.65 (RMSEext of 0.78 and 0.76), respectively. Furthermore, linear and non-linear classification models were developed using the regulatory threshold BCF >2000. A class balanced subset was used to develop classification models which were applied to chemicals not used to create the QSARs. These classification models are characterized by external and internal accuracy up to 84% and 90%, respectively, and sensitivity and specificity up to 90% and 80%, respectively. QSARs presented in this work are validated according to regulatory requirements and their quality is in line with other tools available for the same endpoint and dataset, with the advantage of low complexity and easy application through the software QSAR-ME Profiler. These QSARs can be used as alternatives for, or in combination with, existing models to support bioaccumulation assessment procedures.

Use of a large dataset to develop new models for estimating the sorption of active pharmaceutical ingredients in soils and sediments

Information on the sorption of active pharmaceutical ingredients (APIs) in soils and sediments is needed for assessing the environmental risks of these substances yet these data are unavailable for many APIs in use. Predictive models for estimating sorption could provide a solution. The performance of existing models is, however, often poor and most models do not account for the effects of soil/sediment properties which are known to significantly affect API sorption. Therefore, here, we use a high-quality dataset on the sorption behavior of 54 APIs in 13 soils and sediments to develop new models for estimating sorption coefficients for APIs in soils and sediments using three machine learning approaches (artificial neural network, random forest and support vector machine) and linear regression. A random forest-based model, with chemical and solid descriptors as the input, was the best performing model. Evaluation of this model using an independent sorption dataset from the literature showed that the model was able to predict sorption coefficients of 90% of the test set to within a factor of 10 of the experimental values. This new model could be invaluable in assessing the sorption behavior of molecules that have yet to be tested and in landscape-level risk assessments.

Artificial Intelligence and Machine Learning in Computational Nanotoxicology: Unlocking and Empowering Nanomedicine

Advances in nanomedicine, coupled with novel methods of creating advanced materials at the nanoscale, have opened new perspectives for the development of healthcare and medical products. Special attention must be paid toward safe design approaches for nanomaterial-based products. Recently, artificial intelligence (AI) and machine learning (ML) gifted the computational tool for enhancing and improving the simulation and modeling process for nanotoxicology and nanotherapeutics. In particular, the correlation of in vitro generated pharmacokinetics and pharmacodynamics to in vivo application scenarios is an important step toward the development of safe nanomedicinal products. This review portrays how in vitro and in vivo datasets are used in in silico models to unlock and empower nanomedicine. Physiologically based pharmacokinetic (PBPK) modeling and absorption, distribution, metabolism, and excretion (ADME)-based in silico methods along with dosimetry models as a focus area for nanomedicine are mainly described. The computational OMICS, colloidal particle determination, and algorithms to establish dosimetry for inhalation toxicology, and quantitative structure-activity relationships at nanoscale (nano-QSAR) are revisited. The challenges and opportunities facing the blind spots in nanotoxicology in this computationally dominated era are highlighted as the future to accelerate nanomedicine clinical translation.

A comparison of the persistence, toxicity, and exposure to high-volume natural plant-derived and synthetic pesticides

The immobility of plants exerted evolutionary selection pressures resulting in the production of thousands of chemical substances thought to function as pesticides against predation by insects and animals. More than 10,000 plant-derived compounds have been isolated with the existence of about 100,000 such compounds postulated. In 1990, Ames et al. reported that 99.99% by weight of the pesticides ingested in a normal human diet are derived from natural plant-based sources. This surprising result raised the question as to whether these natural plant pesticides were toxic to humans. These authors examined a relatively small subset of natural pesticides and determined that their tumorigenicity in rodent cancer bioassays was similar to synthetic pesticides. In this analysis, we used standard United States Environmental Protection Agency programs to estimate the toxicity (T.E.S.T. 4.2) and persistence (EPI Suite 4.1) of a series of high-volume synthetic and natural pesticides. On average, synthetic pesticides were more persistent in the environment than were natural pesticides. This result is consistent with cost, time, and logistical constraints under which farmers apply a limited number of applications of pesticides during a crop cycle. Synthetic and natural pesticides are predicted to possess toxicities including mutagenicity and developmental toxicity. Synthetic pesticides are less often mutagenic.

Examining plant uptake and translocation of emerging contaminants using machine learning: Implications to food security

When water and solutes enter the plant root through the epidermis, organic contaminants in solution either cross the root membranes and transport through the vascular pathways to the aerial tissues or accumulate in the plant roots. The accumulation of contaminants in plant roots and edible tissues is measured by root concentration factor (RCF) and fruit concentration factor (FCF). In this paper, 1) a neural network (NN) was applied to model RCF based on physicochemical properties of organic compounds, 2) correlation and significance of physicochemical properties were assessed using statistical analysis, 3) fuzzy logic was used to examine the simultaneous impacts of significant compound properties on RCF and FCF, 4) a clustering algorithm (k-means) was used to identify unique groups and discover hidden relationships within contaminants in various parts of the plants. The physicochemical cutoffs achieved by fuzzy logic for the RCF and the FCF were compared versus the cutoffs for compounds that crossed the plant root membranes and found their way into transpiration stream (measured by transpiration stream concentration factor, TSCF). The NN predicted the RCF with improved accuracy compared to mechanistic models. The analysis indicated that log K_ow, molecular weight, and rotatable bonds are the most important properties for predicting the RCF. These significant compound properties are positively correlated with RCF while they are negatively correlated with TSCF. Comparing the relationships between compound properties in various plant tissues showed that compounds detected in the edible parts have physicochemical cutoffs that are more like the compounds crossing the plant root membranes (into xylem tissues) than the compounds accumulating in the plant roots, with clear relationships to food security. The cluster analysis placed the contaminants into three meaningful groups that were in agreement with the results of fuzzy logic.

QSAR/QSPR models based on quantum chemistry for risk assessment of pesticides according to current European legislation

In Europe, agencies and official organizations involved in the pesticide control such as the EFSA, ECHA, JRC and ECETOC or even the OECD are pointing out that the software tools based on quantitative structure relationship models, i.e. QSAR and QSPR, have a huge potential to improve the pesticide risk assessment process. In this sense, these non-animal test methods can promote the competitiveness of agriculture in this region: the consumer safety is increased with them due to the possibility of perform an overall better risk assessment of the degradation products and metabolites from pesticides. However, the use of theses computational-based (in silico) tools must be much more systematised and harmonised, improving their validation and including case studies to test them. To open databases, incorporating critical data in an orderly manner for building the models, becomes also necessary. Moreover, quantum chemistry through the Density Functional Theory should be promoted as tool for calculation of quantum descriptors, especially for the study of similar compounds with the same carbon skeleton but differing substitution patterns, e.g. isomers.

Methodology of aiQSAR: a group-specific approach to QSAR modelling

Background

Several QSAR methodology developments have shown promise in recent years. These include the consensus approach to generate the final prediction of a model, utilizing new, advanced machine learning algorithms and streamlining, standardization and automation of various QSAR steps. One approach that seems under-explored is at-the-runtime generation of local models specific to individual compounds. This approach was quite likely limited by the computational requirements, but with current increases in processing power and the widespread availability of cluster-computing infrastructure, this limitation is no longer that severe.

Results

We propose a new QSAR methodology: aiQSAR, whose aim is to generate endpoint predictions directly from the input dataset by building an array of local models generated at-the-runtime and specific for each compound in the dataset. The local group of each compound is selected on the basis of fingerprint similarities and the final prediction is calculated by integrating the results of a number of autonomous mathematical models. The method is applicable to regression, binary classification and multi-class classification and was tested on one dataset for each endpoint type: bioconcentration factor (BCF) for regression, Ames test for binary classification and Environmental Protection Agency (EPA) acute rat oral toxicity ranking for multi-class classification. As part of this method, the applicability domain of each prediction is assessed through the applicability domain measure, calculated on the basis of the fingerprint similarities in each local group of compounds.

Conclusions

We outline the methodology for a new QSAR-based predictive tool whose advantages are automation, group-specific approach to modelling and simplicity of execution. Our aim now will be to develop this method into a stand-alone software tool. We hope that eventual adoption of our tool would make QSAR modelling more accessible and transparent. Our methodology could be used as an initial modelling step, to predict new compounds by simply loading the training dataset as an input. Predictions could then be further evaluated and refined either by other tools or through optimization of aiQSAR parameters.

Electronic supplementary material

The online version of this article (10.1186/s13321-019-0350-y) contains supplementary material, which is available to authorized users.

Prediction of bioconcentration factors in fish and invertebrates using machine learning

The application of machine learning has recently gained interest from ecotoxicological fields for its ability to model and predict chemical and/or biological processes, such as the prediction of bioconcentration. However, comparison of different models and the prediction of bioconcentration in invertebrates has not been previously evaluated. A comparison of 24 linear and machine learning models is presented herein for the prediction of bioconcentration in fish and important factors that influenced accumulation identified. R² and root mean square error (RMSE) for the test data (n = 110 cases) ranged from 0.23-0.73 and 0.34-1.20, respectively. Model performance was critically assessed with neural networks and tree-based learners showing the best performance. An optimised 4-layer multi-layer perceptron (14 descriptors) was selected for further testing. The model was applied for cross-species prediction of bioconcentration in a freshwater invertebrate, Gammarus pulex. The model for G. pulex showed good performance with R² of 0.99 and 0.93 for the verification and test data, respectively. Important molecular descriptors determined to influence bioconcentration were molecular mass (MW), octanol-water distribution coefficient (logD), topological polar surface area (TPSA) and number of nitrogen atoms (nN) among others. Modelling of hazard criteria such as PBT, showed potential to replace the need for animal testing. However, the use of machine learning models in the regulatory context has been minimal to date and is critically discussed herein. The movement away from experimental estimations of accumulation to in silico modelling would enable rapid prioritisation of contaminants that may pose a risk to environmental health and the food chain.

Structural alerts for the identification of bioaccumulative compounds

Legislators have included bioaccumulation in the evaluation of chemicals in the framework of the European Registration, Evaluation, Authorization, and Restriction of Chemicals (REACH) regulation. REACH requires information on the bioconcentration factor (BCF), which is a parameter for assessing bioaccumulation and encourages the use of a weight-of-evidence approach, including predictions from quantitative structure-activity relationships (QSARs). This study presents a novel approach, based on structural alerts, to be used as a decision-support system for the identification of substances with bioaccumulation potential. In a regulatory framework, these alerts can be integrated with other sources of information, such as experimental and in silico data, to reduce the uncertainty of the assessment, thereby supporting a weight-of-evidence approach. Moreover, the identified alerts have a direct connection with relevant structural features, thus fostering the applicability and interpretability of the approach. The structural alerts were identified on 779 chemicals annotated for their fish BCF, and the approach was then validated on 278 external molecules. The developed decision-support system allowed identification of 77% of bioaccumulative chemicals and was competitive with more complex QSAR models used in regulatory assessments. The approach is implemented in an easy-to-use workflow, provided free of charge. Integr Environ Assess Manag 2019;15:19-28. © 2018 SETAC.

Acceptable-by-design QSARs to predict the dietary biomagnification of organic chemicals in fish

This work presents the first-time QSAR approach to predict the laboratory-based fish biomagnification factor (BMF) of organic chemicals, to be used as a supporting tool for assessing bioaccumulation at the regulatory level. The developed strategy is based on 2 levels of prediction, with a varying trade-off between interpretability and performance according to the user's needs. We designed our models to be intrinsically acceptable at the regulatory level (in what we defined as "acceptable-by-design" strategy), by (i) complying with OECD principles directly in the approach development phase, (ii) choosing easy-to-apply modeling techniques, (iii) preferring simple descriptors when possible, and (iv) striving to provide data-driven mechanistic insights. Our novel tool has an error comparable to the observed experimental inter- and intraspecies variability and is stable on borderline compounds (root mean square error [RMSE] ranging from RMSE = 0.45 to RMSE = 0.45 log units on test data). Additionally, the models' molecular descriptors are carefully described and interpreted, allowing us to gather additional mechanistic insights into the structural features controlling the dietary bioaccumulation of chemicals in fish. To improve the transparency and promote the application of the model, the data set and the stand alone prediction tool are provided free of charge at https://github.com/grisoniFr/bmf_qsar Integr Environ Assess Manag 2019;15:51-63. © 2018 SETAC.

On the Misleading Use of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msubsup><mml:mi>Q</mml:mi> <mml:mrow><mml:mi>F</mml:mi> <mml:mn>3</mml:mn></mml:mrow> <mml:mn>2</mml:mn></mml:msubsup> </mml:math> for QSAR Model Comparison

Quantitative Structure - Activity Relationship (QSAR) models play a central role in medicinal chemistry, toxicology and computer-assisted molecular design, as well as a support for regulatory decisions and animal testing reduction. Thus, assessing their predictive ability becomes an essential step for any prospective application. Many metrics have been proposed to estimate the model predictive ability of QSARs, which have created confusion on how models should be evaluated and properly compared. Recently, we showed that the metric <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msubsup><mml:mi>Q</mml:mi> <mml:mrow><mml:mi>F</mml:mi> <mml:mn>3</mml:mn></mml:mrow> <mml:mn>2</mml:mn></mml:msubsup> </mml:math> is particularly well-suited for comparing the external predictivity of different models developed on the same training dataset. However, when comparing models developed on different training data, this function becomes inadequate and only dispersion measures like the root-mean-square error (RMSE) should be used. The intent of this work is to provide clarity on the correct and incorrect uses of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msubsup><mml:mi>Q</mml:mi> <mml:mrow><mml:mi>F</mml:mi> <mml:mn>3</mml:mn></mml:mrow> <mml:mn>2</mml:mn></mml:msubsup> </mml:math> , discussing its behavior towards the training data distribution and illustrating some cases in which <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:msubsup><mml:mi>Q</mml:mi> <mml:mrow><mml:mi>F</mml:mi> <mml:mn>3</mml:mn></mml:mrow> <mml:mn>2</mml:mn></mml:msubsup> </mml:math> estimates may be misleading. Hereby, we encourage the usage of measures of dispersions when models trained on different datasets have to be compared and evaluated.

3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction

In this work, we present a new method for predicting complex physical-chemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called three-dimensional reference interaction site model. We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.

Statistical relationship between metabolic decomposition and chemical uptake predicts bioconcentration factor data for diverse chemical exposures

BACKGROUND

A challenge of in vitro to in vivo extrapolation (IVIVE) is to predict the physical state of organisms exposed to chemicals in the environment from in vitro exposure assay data. Although toxicokinetic modeling approaches promise to bridge in vitro screening data with in vivo effects, they are often encumbered by a need for redesign or re-parameterization when applied to different tissues or chemicals.

RESULTS

We demonstrate a parameterization of reverse toxicokinetic (rTK) models developed for the adult zebrafish (Danio rerio) based upon particle swarm optimizations (PSO) of the chemical uptake and degradation rates that predict bioconcentration factors (BCF) for a broad range of chemicals. PSO reveals a relationship between chemical uptake and decomposition parameter values that predicts chemical-specific BCF values with moderate statistical agreement to a limited yet diverse chemical dataset, and all without a need to retrain the model to new data.

CONCLUSIONS

The presented model requires only the octanol-water partitioning ratio to predict BCFs to a fidelity consistent with existing QSAR models. This success begs re-evaluation of the modeling assumptions; specifically, it suggests that chemical uptake into arterial blood may be limited by transport across gill membranes (diffusion) rather than by counter-current flow between gill lamellae (convection). Therefore, more detailed molecular modeling of aquatic respiration may further improve predictive accuracy of the rTK approach.

Environmental risk assessment of selected organic chemicals based on TOC test and QSAR estimation models

Environmental risks of organic chemicals have been greatly determined by their persistence, bioaccumulation, and toxicity (PBT) and physicochemical properties. Major regulations in different countries and regions identify chemicals according to their bioconcentration factor (BCF) and octanol-water partition coefficient (Kow), which frequently displays a substantial correlation with the sediment sorption coefficient (Koc). Half-life or degradability is crucial for the persistence evaluation of chemicals. Quantitative structure activity relationship (QSAR) estimation models are indispensable for predicting environmental fate and health effects in the absence of field- or laboratory-based data. In this study, 39 chemicals of high concern were chosen for half-life testing based on total organic carbon (TOC) degradation, and two widely accepted and highly used QSAR estimation models (i.e., EPI Suite and PBT Profiler) were adopted for environmental risk evaluation. The experimental results and estimated data, as well as the two model-based results were compared, based on the water solubility, Kow, Koc, BCF and half-life. Environmental risk assessment of the selected compounds was achieved by combining experimental data and estimation models. It was concluded that both EPI Suite and PBT Profiler were fairly accurate in measuring the physicochemical properties and degradation half-lives for water, soil, and sediment. However, the half-lives between the experimental and the estimated results were still not absolutely consistent. This suggests deficiencies of the prediction models in some ways, and the necessity to combine the experimental data and predicted results for the evaluation of environmental fate and risks of pollutants.

Predicting the bioconcentration factor through a conformation-independent QSPR study

The ANTARES dataset is a large collection of known and verified experimental bioconcentration factor data, involving 851 highly heterogeneous compounds from which 159 are pesticides. The BCF ANTARES data were used to derive a conformation-independent QSPR model. A large set of 27,017 molecular descriptors was explored, with the main intention of capturing the most relevant structural characteristics affecting the studied property. The structural descriptors were derived with different freeware tools, such as PaDEL, Epi Suite, CORAL, Mold², RECON, and QuBiLs-MAS, and so it was interesting to find out the way that the different descriptor tools complemented each other in order to improve the statistical quality of the established QSPR. The best multivariable linear regression models were found with the Replacement Method variable sub-set selection technique. The proposed QSPR model improves previous reported models of the bioconcentration factor in the present dataset.

Bioaccumulation and ecotoxicity increase during indirect photochemical transformation of polycyclic musk tonalide: A modeling study

Polycyclic musks (PCMs) have recently caused a worldwide environmental concern due to their bioaccumulation potential and ecotoxicological effects. Herein, the OH-initiated indirect photochemical transformation mechanism, environmental fate and ecotoxicity of PCMs (by taking tonalide as an example) were theoretically studied. Results show that tonalide can be degraded readily through OH-addition and H-abstraction pathways, with total rate constants of 6.03 × 10⁹-15.8 × 10⁹ M^-1 s^-1. The OH-addition pathways were dominant at low temperature (<∼287 K), whereas H-abstraction was the dominant pathway at high temperature. Further, the bioconcentration factors (BCF) and aquatic toxicities to fish of all transformation products from H-abstraction pathways were smaller than tonalide. In contrast, these values of most intermediates from OH-addition pathways were up to 8 times higher than tonalide. Particularly, the resultant phenolic product PC1 had a BCF of 5590 L/kg wet-wt, which exceeds the cutoff criterion set for the typically persistent organic pollutants as critically bioaccumulative. Notably, PC1 would mainly be produced under anaerobic aquatic conditions at low temperatures. Therefore, particular attention should be paid to the indirect photochemical products and parental PCMs, particularly the intermediates from OH-addition pathway.

Integrated in silico strategy for PBT assessment and prioritization under REACH

Chemicals may persist in the environment, bioaccumulate and be toxic for humans and wildlife, posing great concern. These three properties, persistence (P), bioaccumulation (B), and toxicity (T) are the key targets of the PBT-hazard assessment. The European regulation for the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) requires assessment of PBT-properties for all chemicals that are produced or imported in Europe in amounts exceeding 10 tonnes per year, checking whether the criteria set out in REACH Annex XIII are met, so the substance should therefore be considered to have properties of very high concern. Considering how many substances can fall under the REACH regulation, there is a pressing need for new strategies to identify and screen large numbers fast and inexpensively. An efficient non-testing screening approach to identify PBT candidates is necessary, as a valuable alternative to money- and time-consuming laboratory tests and a good start for prioritization since few tools exist (e.g. the PBT profiler developed by US EPA). The aim of this work was to offer a conceptual scheme for identifying and prioritizing chemicals for further assessment and if appropriate further testing, based on their PBT-potential, using a non-testing screening approach. We integrated in silico models (using existing and developing new ones) in a final algorithm for screening and ranking PBT-potential, which uses experimental and predicted values as well as associated uncertainties. The Multi-Criteria Decision-Making (MCDM) theory was used to integrate the different values. Then we compiled a new set of data containing known PBT and non-PBT substances, in order to check how well our approach clearly differentiated compounds labeled as PBT from those labeled as non-PBT. This indicated that the integrated model distinguished between PBT from non-PBT compounds.

An ensemble model of QSAR tools for regulatory risk assessment

Quantitative structure activity relationships (QSARs) are theoretical models that relate a quantitative measure of chemical structure to a physical property or a biological effect. QSAR predictions can be used for chemical risk assessment for protection of human and environmental health, which makes them interesting to regulators, especially in the absence of experimental data. For compatibility with regulatory use, QSAR models should be transparent, reproducible and optimized to minimize the number of false negatives. In silico QSAR tools are gaining wide acceptance as a faster alternative to otherwise time-consuming clinical and animal testing methods. However, different QSAR tools often make conflicting predictions for a given chemical and may also vary in their predictive performance across different chemical datasets. In a regulatory context, conflicting predictions raise interpretation, validation and adequacy concerns. To address these concerns, ensemble learning techniques in the machine learning paradigm can be used to integrate predictions from multiple tools. By leveraging various underlying QSAR algorithms and training datasets, the resulting consensus prediction should yield better overall predictive ability. We present a novel ensemble QSAR model using Bayesian classification. The model allows for varying a cut-off parameter that allows for a selection in the desirable trade-off between model sensitivity and specificity. The predictive performance of the ensemble model is compared with four in silico tools (Toxtree, Lazar, OECD Toolbox, and Danish QSAR) to predict carcinogenicity for a dataset of air toxins (332 chemicals) and a subset of the gold carcinogenic potency database (480 chemicals). Leave-one-out cross validation results show that the ensemble model achieves the best trade-off between sensitivity and specificity (accuracy: 83.8 % and 80.4 %, and balanced accuracy: 80.6 % and 80.8 %) and highest inter-rater agreement [kappa (κ): 0.63 and 0.62] for both the datasets. The ROC curves demonstrate the utility of the cut-off feature in the predictive ability of the ensemble model. This feature provides an additional control to the regulators in grading a chemical based on the severity of the toxic endpoint under study.

AutoQSAR: an automated machine learning tool for best-practice quantitative structure-activity relationship modeling

AIM

We introduce AutoQSAR, an automated machine-learning application to build, validate and deploy quantitative structure-activity relationship (QSAR) models.

METHODOLOGY/RESULTS

The process of descriptor generation, feature selection and the creation of a large number of QSAR models has been automated into a single workflow within AutoQSAR. The models are built using a variety of machine-learning methods, and each model is scored using a novel approach. Effectiveness of the method is demonstrated through comparison with literature QSAR models using identical datasets for six end points: protein-ligand binding affinity, solubility, blood-brain barrier permeability, carcinogenicity, mutagenicity and bioaccumulation in fish.

CONCLUSION

AutoQSAR demonstrates similar or better predictive performance as compared with published results for four of the six endpoints while requiring minimal human time and expertise.

The origin and evolution of assessment criteria for persistent, bioaccumulative and toxic (PBT) chemicals and persistent organic pollutants (POPs)

General public concern over the effects of persistent chemicals began in the early 1960s. Since then, significant scientific advances have increased our understanding of persistent, bioaccumulative, and toxic (PBT) chemicals and the properties and processes that influence their fates in, and adverse effects on, human health and the environment. In addition to the scientific advances, a number of legislations and agreements for national, international, and global identification and control of PBT chemicals have been adopted. However, some of the rationales and thoughts that were relied upon when the first criteria were developed to identify and categorize PBT chemicals and then POPs (persistent organic pollutants) have not been carried forward. Criteria have been based upon available data of neutral hydrophobic substances as reference chemicals, derived under laboratory conditions. They evolved over the last decades due to the diversification of the protection aims under various national regulatory frameworks and international agreements, advances in methods for estimation of physical/chemical properties, and the identification of chemicals which are non-traditional POPs. Criteria are not defined purely by science; they also are subject to the aims of policy. This paper offers a historical perspective on the development of criteria for PBT chemicals and POPs. It also offers suggestions for rationalization of protection goals, describes some emerging procedures for identification of compounds of concern, and proposes information that needs to be considered when applying criteria to screening and/or evaluation of new chemicals.

Expert QSAR system for predicting the bioconcentration factor under the REACH regulation

Expert systems are a rational integration of several models that generally aim to exploit their advantages and overcome their drawbacks. This work is founded on our previously published Quantitative Structure-Activity Relationship (QSAR) classification scheme, which detects compounds whose Bioconcentration Factor (BCF) is (1) well predicted by the octanol-water partition coefficient (KOW), (2) underestimated by KOW or (3) overestimated by KOW. The classification scheme served as the starting point to identify and combine the best BCF model for each class among three VEGA models and one KOW-based equation. The rationalized model integration showed stability and surprising performance on unknown data when compared with benchmark BCF models. Model simplicity, transparency and mechanistic interpretation were fostered in order to allow for its application and acceptance within the REACH framework.

Occurrence, removal and risk assessment of pharmaceutical and personal care products (PPCPs) in an advanced drinking water treatment plant (ADWTP) around Taihu Lake in China

The occurrence and removal of 39 selected pharmaceutical and personal care products (PPCPs) were investigated in an advanced drinking water treatment plant (ADWTP) around Taihu Lake. Fourteen of 39 targeted pharmaceuticals were detected in the raw water. After a series of purification processes, only indomethacin, caffeine and sulfamethoxazole were found in effluent, albeit at concentrations less than 2 ng L(-1). The results of principal component analysis suggested that three main purification processes, oxidation, coagulation combined with sedimentation and filtration combined with bio-degradation, influenced the removal performance of PPCPs. The ecotoxicological and human health risk assessment confirmed that drugs detected in effluent posed no potential toxicity and also suggested that two PPCPs (roxithromycin and sulfamethoxazole), especially sulfamethoxazole, should be seriously considered as candidates for regulatory monitoring and prioritization. Finally, the correlation between removal efficiency and risk quotient indicated that uniform removal efficiency for all PPCPs may not reflect an equal risk control in the ADWTP.

Comparison between bioconcentration factor (BCF) data provided by industry to the European Chemicals Agency (ECHA) and data derived from QSAR models

The bioconcentration factor (BCF) is the ratio of the concentration of a chemical in an organism to the concentration in the surrounding environment at steady state. It is a valuable indicator of the bioaccumulation potential of a substance. BCF is an essential environmental property required for regulatory purposes within the Registration, Evaluation, Authorization and restriction of Chemicals (REACH) and Globally Harmonized System (GHS) regulations. In silico models for predicting BCF can facilitate the risk assessment for aquatic toxicology and reduce the cost and number of animals used. The aim of the present study was to examine the correlation of BCF data derived from the dossiers of registered chemicals submitted to the European Chemical Agency (ECHA) with the results of a battery of Quantitative Structure-Activity Relationship (QSAR). After data pruning, statistical analysis was performed using the predictions of the selected models. Results in terms of R(2) had low rating around 0.5 for the pruned dataset. The use of the model applicability domain index (ADI) led to an improvement of the performance for compounds falling within it. The variability of the experimental data and the use of different parameters to define the applicability domain can influence the performance of each model. All available information should be adapted to the requirements of the regulation to obtain a safe decision.

Integrating QSAR and read-across for environmental assessment

Read-across and QSAR have different traditions and drawbacks. We address here two main questions: (1) How do we solve the issue of the subjectivity in the evaluation of data and results, which may be particularly critical for read-across, but may have a role also for the QSAR assessment? (2) How do we take advantage of the results of both approaches to support each other? The QSAR model starts from the training set. The presence of similar chemicals with property values close to that predicted can support the result. The approach in read-across is the opposite. The assessment is focused on the few substances similar to the target. The data quality of the similar chemicals is fundamental. A risk is poor standardization in the definition of 'similarity', because different approaches may be applied. Inspired by the principles of high transparency and reproducibility, a new program for read-across, called ToxRead, has been developed and made freely available ( www.toxgate.eu ). The output of ToxRead can be compared and integrated with the output of QSAR, within a weight-of-evidence strategy. We discuss the evaluation and integration of ToxRead and QSAR with examples of the assessment of bioconcentration factors of chemicals.

QSAR models for bioconcentration: is the increase in the complexity justified by more accurate predictions?

This study compares nine QSAR models for the prediction of BCF on fish: four KOW based models (Veith, Mackay, Bintein and TGD equations) and five complex models (EPI Suite BCFBAF, VEGA CAESAR, VEGA Meylan, VEGA Read-across and VEGA consensus). The aim is to test if increasing complexity leads to predictions that are more accurate than those based only on KOW are. To this end, experimental BCF data for 1056 compounds, along with experimental and predicted KOW values, were collected and used for the comparison. A particular focus has been placed on compounds for which metabolism, elimination and specific interactions with tissues can be hypothesized. VEGA Read-across improved global predictions with respect to the KOW based models and resulted to be a good approach to take into account metabolism and interactions with tissues. For the other complex models, several drawbacks were highlighted. Finally, for different classes of compounds (i.e. Perfluorinated Compounds, Organophosphorous Compounds, Synthetic Pyrethroids and Polychlorinated Biphenyls) results confirmed the mechanistic interpretation of the processes involved in their bioconcentration.

Evaluation and comparison of benchmark QSAR models to predict a relevant REACH endpoint: The bioconcentration factor (BCF)

The bioconcentration factor (BCF) is an important bioaccumulation hazard assessment metric in many regulatory contexts. Its assessment is required by the REACH regulation (Registration, Evaluation, Authorization and Restriction of Chemicals) and by CLP (Classification, Labeling and Packaging). We challenged nine well-known and widely used BCF QSAR models against 851 compounds stored in an ad-hoc created database. The goodness of the regression analysis was assessed by considering the determination coefficient (R(2)) and the Root Mean Square Error (RMSE); Cooper's statistics and Matthew's Correlation Coefficient (MCC) were calculated for all the thresholds relevant for regulatory purposes (i.e. 100L/kg for Chemical Safety Assessment; 500L/kg for Classification and Labeling; 2000 and 5000L/kg for Persistent, Bioaccumulative and Toxic (PBT) and very Persistent, very Bioaccumulative (vPvB) assessment) to assess the classification, with particular attention to the models' ability to control the occurrence of false negatives. As a first step, statistical analysis was performed for the predictions of the entire dataset; R(2)>0.70 was obtained using CORAL, T.E.S.T. and EPISuite Arnot-Gobas models. As classifiers, ACD and logP-based equations were the best in terms of sensitivity, ranging from 0.75 to 0.94. External compound predictions were carried out for the models that had their own training sets. CORAL model returned the best performance (R(2)ext=0.59), followed by the EPISuite Meylan model (R(2)ext=0.58). The latter gave also the highest sensitivity on external compounds with values from 0.55 to 0.85, depending on the thresholds. Statistics were also compiled for compounds falling into the models Applicability Domain (AD), giving better performances. In this respect, VEGA CAESAR was the best model in terms of regression (R(2)=0.94) and classification (average sensitivity>0.80). This model also showed the best regression (R(2)=0.85) and sensitivity (average>0.70) for new compounds in the AD but not present in the training set. However, no single optimal model exists and, thus, it would be wise a case-by-case assessment. Yet, integrating the wealth of information from multiple models remains the winner approach.

Use of artificial neural networks in the QSAR prediction of physicochemical properties and toxicities for REACH legislation

With the introduction of the REACH legislation in the European Union, there is a requirement for property and toxicity data on chemicals produced in or imported into the EU at levels of 1 tonne/year or more. This has meant an increase in the in silico prediction of such data. One of the chief predictive approaches is QSAR (quantitative structure-activity relationships), which is widely used in many fields. A QSAR approach that is increasingly being used is that of artificial neural networks (ANNs), and this chapter discusses its application to the range of physicochemical properties and toxicities required by REACH. ANNs generally outperform the main QSAR approach of multiple linear regression (MLR), although other approaches such as support vector machines sometimes outperform ANNs. Most ANN QSARs reported to date comply with only two of the five OECD Guidelines for the Validation of (Q)SARs.

Comparative performance of descriptors in a multiple linear and Kriging models: a case study on the acute toxicity of organic chemicals to algae

This study presents quantitative structure-toxicity relationship (QSTR) models on the toxicity of 91 organic compounds to Chlorella vulgaris using multiple linear regression (MLR) and Kriging techniques. The molecular descriptors were calculated using SPARTAN and DRAGON programs, and descriptor selection was made by "all subset" method available in the QSARINS software. MLR and Kriging models developed with the same descriptors were compared. In addition to these models, Kriging method was used for descriptor selection, and model development. The selected descriptors showed the importance of hydrophobicity, molecular weight and atomic ionization state in describing the toxicity of a diverse set of chemicals to C. vulgaris. A QSTR model should be associated with appropriate measures of goodness-of-fit, robustness, and predictivity in order to be used for regulatory purpose. Therefore, while the internal performances (goodness-of-fit and robustness) of the models were determined by using a training set, the predictive abilities of the models were determined by using a test set. The results of the study showed that while MLR method is easier to apply, the Kriging method was more successful in predicting toxicity.

Synergistic effects of perfluoroalkyl acids mixtures with J-shaped concentration-responses on viability of a human liver cell line

Some perfluoroalkyl acids (PFAAs) are highly persistent and bioaccumulative, resulting in their broad coexisting distribution in humans and the environment. Our aim was to investigate the individual and joint effects of PFAAs on cellular viability of a human liver cell line (HL-7702) using the MTT assay. Equipartition ray design and equivalent-effect concentration ratio (EECR) mixtures were used to investigate the binary and multiple effects of PFAAs, respectively. All tested PFAAs mixtures and the individuals (except perfluorododecanoic acid (PFDoDA) and perfluorotetradecanoic acid (PFTeDA)) showed obvious non-monotonic J-shaped concentration-response curves (CRC) on HL-7702. The inhibitory effect of individual PFAAs increased with the elongation of the carbon chain and was dominated by their molecular volume. The three binary mixtures (PFOA/S, PFHxA/S and PFBA/S) showed that synergistic effects occurred under effective inhibitory concentrations (IC) of IC0, IC10, and IC50 in mixtures, while for IC-20 the synergistic effect only occurred under higher PFSA proportion in mixtures. Furthermore, EECR mixtures of the nine individual PFAAs with J-shaped CRC also showed synergistic effects. However, mixtures of the eleven individual PFAAs including those with S-shaped CRC resulted in partial addition effects on HL-7702. Our results indicated that the individual stimulatory responses of HL-7702 to PFAA may produce adverse effects in mixtures at relevant dose levels.

Modeling bioconcentration factor (BCF) using mechanistically interpretable descriptors computed from open source tool "PaDEL-Descriptor"

Predictive regression-based models for bioconcentration factor (BCF) have been developed using mechanistically interpretable descriptors computed from open source tool PaDEL-Descriptor ( http://padel.nus.edu.sg/software/padeldescriptor/ ). A data set of 522 diverse chemicals has been used for this modeling study, and extended topochemical atom (ETA) indices developed by the present authors' group were chosen as the descriptors. Due to the importance of lipohilicity in modeling BCF, XLogP (computed partition coefficient) was also tried as an additional descriptor. Genetic function approximation followed by multiple linear regression algorithm was applied to select descriptors, and subsequent partial least squares analyses were performed to establish mathematical equations for BCF prediction. The model generated from only ETA indices shows importance of seven descriptors in model development, while the model generated from ETA descriptors along with XlogP shows importance of four descriptors in model development. In general, BCF depends on lipophilicity, presence of heteroatoms, presence of halogens, fused ring system, hydrogen bonding groups, etc. The developed models show excellent statistical qualities and predictive ability. The developed models were used also for prediction of an external data set available from the literature, and good quality of predictions (R (2) pred = 0.812 and 0.826) was demonstrated. Thus, BCF can be predicted using ETA and XlogP descriptors calculated from open source PaDEL-Descriptor software in the context of aquatic chemical toxicity management.

A reductionist mechanistic model for bioconcentration of neutral and weakly polar organic compounds in fish

The bioconcentration factor (BCF) of neutral and weakly polar organic chemicals in fish is modeled using independently calibrated models of chemical partitioning (freely dissolved fraction of chemical in the aqueous phase [φsys ] and wet-weight fish-water partition coefficient [KFW ]), respiratory exchange (respiratory update rate constant [k1 ], and respiratory elimination rate constant [k2  = k1 /KFW ]), and biotransformation (whole-body biotransformation rate constant [kM ]) as BCF = φsys KFW /(1 + kM /k2 ). Existing k1 models tend to overestimate for chemicals with log KOW  < 3.5, which constituted 30% to 50% of the examined chemicals. A revised k1 model covering a wider log KOW range (0-8.5) is presented k1  = (5.46 × 10(-6) MW + 0.261/KOW )(-1) , where MW is the molecular weight. The biotransformation rate constant kM is modeled using biota internal partitioning and Abraham parameters as reactivity descriptors. The reductionist model was tested using 3 different BCF data sets (US Environmental Protection Agency's Estimation Programs Interface [EPI], n = 548; Hertfordshire, n = 210; Arnot-Gobas, n = 1855) and compared with the following 3 state-of-the-art models: 1) the EPI Suite BCFBAF module, 2) the European Commision's Computer Assisted Evaluation of industrial chemical Substances According to Regulations (CAESAR), and 3) the EPI/Arnot mechanistic kinetic model. The reductionist model performed comparably with the alternative models (root mean square errors [RMSEs] = 0.72-0.77), with only 5 fitting parameters and no training against experimental BCFs. Respiratory elimination and biotransformation dominate the total depuration (i.e., [k2  + kM ]/kT  ≥ 0.8) for approximately 98% of the data entries, thus validating the reductionist approximation. Mechanistic models provide greater insights into bioaccumulation and are more sensitive to biological variation. All three BCF data sets and relevant properties and checkpoint values necessary for reproducing predictions of the reductionist model have been documented. The present study shows that a streamlined mechanistic model of BCF is possible for assessment purposes.

Biotransformation model of neutral and weakly polar organic compounds in fish incorporating internal partitioning

A model for whole-body in vivo biotransformation of neutral and weakly polar organic chemicals in fish is presented. It considers internal chemical partitioning and uses Abraham solvation parameters as reactivity descriptors. It assumes that only chemicals freely dissolved in the body fluid may bind with enzymes and subsequently undergo biotransformation reactions. Consequently, the whole-body biotransformation rate of a chemical is retarded by the extent of its distribution in different biological compartments. Using a randomly generated training set (n = 64), the biotransformation model is found to be: log (HLφfish ) = 2.2 (±0.3)B - 2.1 (±0.2)V - 0.6 (±0.3) (root mean square error of prediction [RMSE] = 0.71), where HL is the whole-body biotransformation half-life in days, φfish is the freely dissolved fraction in body fluid, and B and V are the chemical's H-bond acceptance capacity and molecular volume. Abraham-type linear free energy equations were also developed for lipid-water (Klipidw ) and protein-water (Kprotw ) partition coefficients needed for the computation of φfish from independent determinations. These were found to be 1) log Klipidw  = 0.77E - 1.10S - 0.47A - 3.52B + 3.37V + 0.84 (in Lwat /kglipid ; n = 248, RMSE = 0.57) and 2) log Kprotw  = 0.74E - 0.37S - 0.13A - 1.37B + 1.06V - 0.88 (in Lwat /kgprot ; n = 69, RMSE = 0.38), where E, S, and A quantify dispersive/polarization, dipolar, and H-bond-donating interactions, respectively. The biotransformation model performs well in the validation of HL (n = 424, RMSE = 0.71). The predicted rate constants do not exceed the transport limit due to circulatory flow. Furthermore, the model adequately captures variation in biotransformation rate between chemicals with varying log octanol-water partitioning coefficient, B, and V and exhibits high degree of independence from the choice of training chemicals. The present study suggests a new framework for modeling chemical reactivity in biological systems.

From data point timelines to a well curated data set, data mining of experimental data and chemical structure data from scientific articles, problems and possible solutions

The scientific literature is important source of experimental and chemical structure data. Very often this data has been harvested into smaller or bigger data collections leaving the data quality and curation issues on shoulders of users. The current research presents a systematic and reproducible workflow for collecting series of data points from scientific literature and assembling a database that is suitable for the purposes of high quality modelling and decision support. The quality assurance aspect of the workflow is concerned with the curation of both chemical structures and associated toxicity values at (1) single data point level and (2) collection of data points level. The assembly of a database employs a novel "timeline" approach. The workflow is implemented as a software solution and its applicability is demonstrated on the example of the Tetrahymena pyriformis acute aquatic toxicity endpoint. A literature collection of 86 primary publications for T. pyriformis was found to contain 2,072 chemical compounds and 2,498 unique toxicity values, which divide into 2,440 numerical and 58 textual values. Every chemical compound was assigned to a preferred toxicity value. Examples for most common chemical and toxicological data curation scenarios are discussed.

Integration of QSAR models for bioconcentration suitable for REACH

QSAR (Quantitative Structure Activity Relationship) models can be a valuable alternative method to replace or reduce animal test required by REACH. In particular, some endpoints such as bioconcentration factor (BCF) are easier to predict and many useful models have been already developed. In this paper we describe how to integrate two popular BCF models to obtain more reliable predictions. In particular, the herein presented integrated model relies on the predictions of two among the most used BCF models (CAESAR and Meylan), together with the Applicability Domain Index (ADI) provided by the software VEGA. Using a set of simple rules, the integrated model selects the most reliable and conservative predictions and discards possible outliers. In this way, for the prediction of the 851 compounds included in the ANTARES BCF dataset, the integrated model discloses a R(2) (coefficient of determination) of 0.80, a RMSE (Root Mean Square Error) of 0.61 log units and a sensitivity of 76%, with a considerable improvement in respect to the CAESAR (R(2)=0.63; RMSE=0.84 log units; sensitivity 55%) and Meylan (R(2)=0.66; RMSE=0.77 log units; sensitivity 65%) without discarding too many predictions (118 out of 851). Importantly, considering solely the compounds within the new integrated ADI, the R(2) increased to 0.92, and the sensitivity to 85%, with a RMSE of 0.44 log units. Finally, the use of properly set safety thresholds applied for monitoring the so called "suspicious" compounds, which are those chemicals predicted in proximity of the border normally accepted to discern non-bioaccumulative from bioaccumulative substances, permitted to obtain an integrated model with sensitivity equal to 100%.

Meta-analysis of toxicity and teratogenicity of 133 chemicals from zebrafish developmental toxicity studies

Zebrafish developmental toxicity testing is an emerging field, which faces considerable challenges regarding data meta-analysis and the establishment of standardized test protocols. Here, we present an initial correlation study on toxicity of 133 chemicals based on data in the literature to ascertain predictive developmental toxicity endpoints. We found that the physical properties of chemicals (BCF or logP) did not fully predict lethality or developmental outcomes. Instead, individual outcomes such as pericardial edema and yolk sac edema were more reliable indicators of developmental toxicity. In addition, we ranked the chemicals based on toxicity with the Toxicological Priority Index (ToxPi) program and via a teratogenic ratio, and found that perfluorooctane sulfonate (PFOS) had the highest ToxPi score, triphenyltin acetate had the highest average ToxPi score (corrected for missing data and having more than 4 outcomes), and N-methyl-dithiocarbamate had the highest teratogenic ratio.

Defining a novel k-nearest neighbours approach to assess the applicability domain of a QSAR model for reliable predictions

Background

With the growing popularity of using QSAR predictions towards regulatory purposes, such predictive models are now required to be strictly validated, an essential feature of which is to have the model’s Applicability Domain (AD) defined clearly. Although in recent years several different approaches have been proposed to address this goal, no optimal approach to define the model’s AD has yet been recognized.

Results

This study proposes a novel descriptor-based AD method which accounts for the data distribution and exploits k-Nearest Neighbours (kNN) principle to derive a heuristic decision rule. The proposed method is a three-stage procedure to address several key aspects relevant in judging the reliability of QSAR predictions. Inspired from the adaptive kernel method for probability density function estimation, the first stage of the approach defines a pattern of thresholds corresponding to the various training samples and these thresholds are later used to derive the decision rule. Criterion deciding if a given test sample will be retained within the AD is defined in the second stage of the approach. Finally, the last stage tries reflecting upon the reliability in derived results taking model statistics and prediction error into account.

Conclusions

The proposed approach addressed a novel strategy that integrated the kNN principle to define the AD of QSAR models. Relevant features that characterize the proposed AD approach include: a) adaptability to local density of samples, useful when the underlying multivariate distribution is asymmetric, with wide regions of low data density; b) unlike several kernel density estimators (KDE), effectiveness also in high-dimensional spaces; c) low sensitivity to the smoothing parameter k; and d) versatility to implement various distances measures. The results derived on a case study provided a clear understanding of how the approach works and defines the model’s AD for reliable predictions.

QSAR workbench: automating QSAR modeling to drive compound design

We describe the QSAR Workbench, a system for the building and analysis of QSAR models. The system is built around the Pipeline Pilot workflow tool and provides access to a variety of model building algorithms for both continuous and categorical data. Traditionally models are built on a one by one basis and fully exploring the model space of algorithms and descriptor subsets is a time consuming basis. The QSAR Workbench provides a framework to allow for multiple models to be built over a number of modeling algorithms, descriptor combinations and data splits (training and test sets). Methods to analyze and compare models are provided, enabling the user to select the most appropriate model. The Workbench provides a consistent set of routines for data preparation and chemistry normalization that are also applied for predictions. The Workbench provides a large degree of automation with the ability to publish preconfigured model building workflows for a variety of problem domains, whilst providing experienced users full access to the underlying parameterization if required. Methods are provided to allow for publication of selected models as web services, thus providing integration with the chemistry desktop. We describe the design and implementation of the QSAR Workbench and demonstrate its utility through application to two public domain datasets.

Using conditional inference trees and random forests to predict the bioaccumulation potential of organic chemicals

The present study presents a data-oriented, tiered approach to assessing the bioaccumulation potential of chemicals according to the European chemicals regulation on Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH). The authors compiled data for eight physicochemical descriptors (partition coefficients, degradation half-lives, polarity, and so forth) for a set of 713 organic chemicals for which experimental values of the bioconcentration factor (BCF) are available. The authors employed supervised machine learning methods (conditional inference trees and random forests) to derive relationships between the physicochemical descriptors and the BCF values. In a first tier, the authors established rules for classifying a chemical as bioaccumulative (B) or nonbioaccumulative (non-B). In a second tier, the authors developed a new tool for estimating numerical BCF values. For both cases the optimal set of relevant descriptors was determined; these are biotransformation half-life and octanol-water distribution coefficient (log D) for the classification rules and log D, biotransformation half-life, and topological polar surface area for the BCF estimation tool. The uncertainty of the BCF estimates obtained with the new estimation tool was quantified by comparing the estimated and experimental BCF values of the 713 chemicals. Comparison with existing BCF estimation methods indicates that the performance of this new BCF estimation tool is at least as high as that of existing methods. The authors recommend the present study's classification rules and BCF estimation tool for a consensus application in combination with existing BCF estimation methods.

Comparative analysis of local and consensus quantitative structure-activity relationship approaches for the prediction of bioconcentration factor

Quantitative structure-activity relationships (QSARs) are broadly classified as global or local, depending on their molecular constitution. Global models use large and diverse training sets covering a wide range of chemical space. Local models focus on smaller structurally or chemically similar subsets that are conventionally selected by human experts or alternatively using clustering analysis. The current study focuses on the comparative analysis of different clustering algorithms (expectation-maximization, K-means and hierarchical) for seven different descriptor sets as structural characteristics and two rule-based approaches to select subsets for designing local QSAR models. A total of 111 local QSAR models are developed for predicting bioconcentration factor. Predictions from local models were compared with corresponding predictions from the global model. The comparison of coefficients of determination (r(2)) and standard deviations for local models with similar subsets from the global model show improved prediction quality in 97% of cases. The descriptor content of derived QSARs is discussed and analyzed. Local QSAR models were further consolidated within the framework of consensus approach. All different consensus approaches increased performance over the global and local models. The consensus approach reduced the number of strongly deviating predictions by evening out prediction errors, which were produced by some local QSARs.