Predicting skin sensitisation using a decision tree integrated testing strategy with an in silico model and in chemico/in vitro assays

Reconstructed human epidermis-based testing strategy of skin sensitization potential and potency classification using epidermal sensitization assay and in silico data

The hazards and potency of skin sensitizers are traditionally determined using animal tests such as the local lymph node assay (LLNA); however, significant progress has been made in the development of non-animal test methods addressing the first three mechanistic key events of adverse outcome pathway in skin sensitization. We developed the epidermal sensitization assay (EpiSensA), which is a reconstructed human epidermis-based assay, by measuring four genes related to critical keratinocyte responses during skin sensitization. Four in vitro skin sensitization test methods (EpiSensA, direct peptide reactivity assay [DPRA], KeratinoSens™, and human cell line activation test [h-CLAT]) were systematically evaluated using 136 chemicals including lipophilic chemicals and pre/pro-haptens, which may be related to assay-specific limitations. The constructed database included existing and newly generated data. The EpiSensA showed a broader applicability domain and predicted the hazards with 82.4% and 78.8% accuracy than LLNA and human data. The EpiSensA could detect 76 out of 88 sensitizers at lower concentrations than the LLNA, indicating that the EpiSensA has higher sensitivity for the detection of minor sensitizing constituents. These results confirmed the potential use of the EpiSensA in evaluating a mixture of unknown compositions that can be evaluated by animal tests. To combine different information sources, the reconstructed human epidermis-based testing strategy (RTS) was developed based on weighted multiple information from the EpiSensA and TImes MEtabolism Simulator platform for predicting Skin Sensitization (TIMES-SS; RTSv1) or Organization for Economic Cooperation and Development (OECD) QSAR Toolbox automated workflow (RTSv2). The predictivities of the hazards and Globally Harmonized System (GHS) subcategories were equal to or better than the defined approaches (2 out of 3, integrated testing strategy [ITS]v1, and ITSv2) adopted as OECD Guideline 497.

Review of machine learning and deep learning models for toxicity prediction

The ever-increasing number of chemicals has raised public concerns due to their adverse effects on human health and the environment. To protect public health and the environment, it is critical to assess the toxicity of these chemicals. Traditional in vitro and in vivo toxicity assays are complicated, costly, and time-consuming and may face ethical issues. These constraints raise the need for alternative methods for assessing the toxicity of chemicals. Recently, due to the advancement of machine learning algorithms and the increase in computational power, many toxicity prediction models have been developed using various machine learning and deep learning algorithms such as support vector machine, random forest, k-nearest neighbors, ensemble learning, and deep neural network. This review summarizes the machine learning- and deep learning-based toxicity prediction models developed in recent years. Support vector machine and random forest are the most popular machine learning algorithms, and hepatotoxicity, cardiotoxicity, and carcinogenicity are the frequently modeled toxicity endpoints in predictive toxicology. It is known that datasets impact model performance. The quality of datasets used in the development of toxicity prediction models using machine learning and deep learning is vital to the performance of the developed models. The different toxicity assignments for the same chemicals among different datasets of the same type of toxicity have been observed, indicating benchmarking datasets is needed for developing reliable toxicity prediction models using machine learning and deep learning algorithms. This review provides insights into current machine learning models in predictive toxicology, which are expected to promote the development and application of toxicity prediction models in the future.

The Good, The Bad, and The Perplexing: Structural Alerts and Read-Across for Predicting Skin Sensitization Using Human Data

In our earlier work (Golden et al., 2021), we showed 70-80% accuracies for several skin sensitization computational tools using human data. Here, we expanded the data set using the NICEATM human skin sensitization database to create a final data set of 1355 discrete chemicals (largely negative, ∼70%). Using this expanded data set, we analyzed model performance and evaluated mispredictions using Toxtree (v 3.1.0), OECD QSAR Toolbox (v 4.5), VEGA's (1.2.0 BETA) CAESAR (v 2.1.7), and a k-nearest-neighbor (kNN) classification approach. We show that the accuracy on this data set was lower than previous estimates, with balanced accuracies being 63% and 65% for Toxtree and OECD QSAR Toolbox, respectively, 46% for VEGA, and 59% for a kNN approach, with the lower accuracy likely due to the higher percentage of nonsensitizing chemicals. Two hundred eighty seven chemicals were mispredicted by both Toxtree and OECD QSAR Toolbox, which was approximately 20% of the entire data set, and 84% of these were false positives. The absence or presence of metabolic simulation in OECD QSAR Toolbox made no overall difference. While Toxtree is known for overpredicting, 60% of the chemicals in the data set had no alert for skin sensitization, and a substantial number of these chemicals were in fact sensitizers, pointing to sensitization mechanisms not recognized by Toxtree. Interestingly, we observed that chemicals with more than one Toxtree alert were more likely to be nonsensitizers. Finally, a kNN approach tended to mispredict different chemicals than either OECD QSAR Toolbox or Toxtree, suggesting that there was additional information to be garnered from a kNN approach. Overall, the results demonstrate that while there is merit in structural alerts as well as QSAR or read-across approaches (perhaps even more so in their combination), additional improvement will require a more nuanced understanding of mechanisms of skin sensitization.

The activity of methacrylate esters in skin sensitisation test methods II. A review of complementary and additional analyses

Allergic contact dermatitis is an important occupational health issue, and there is a need to identify accurately those chemicals that have the potential to induce skin sensitisation. Hazard identification was performed initially using animal (guinea pig and mouse) models. More recently, as a result of the drive towards non-animal methods, alternative in vitro and in silico approaches have been developed. Some of these new in vitro methods have been formally validated and have been assigned OECD Test Guideline status. The performance of some of these recently developed in vitro methods, and of 2 quantitative structure-activity relationships (QSAR) approaches, with a series of methacrylate esters has been reviewed and reported previously. In this article that first review has been extended further with additional data and complementary analyses. Results obtained using in vitro methods (Direct Peptide Reactivity Assay, DPRA; ARE-Nrf2 luciferase test methods, KeratinoSens and LuSens; Epidermal Sensitisation Assay, EpiSensA; human Cell Line Activation Test, h-CLAT, and the myeloid U937 Skin Sensitisation test, U-SENS), and 2 QSAR approaches (DEREK™-nexus and TIMES-SS), with 11 methacrylate esters and methacrylic acid are reported here, and compared with existing data from the guinea pig maximisation test and the local lymph node assay. With this series of chemicals it was found that some in vitro tests (DPRA and ARE-Nrf2 luciferase) performed well in comparison with animal test results and available human skin sensitisation data. Other in vitro tests (EpiSensA and h-CLAT) proved rather more problematic. Results with DEREK™-nexus and TIMES-SS failed to reflect accurately the skin sensitisation potential of the methacrylate esters. The implications for assessment of skin sensitising activity are discussed.

Application of Spectro-DPRA, KeratinoSens™ and h-CLAT to estimation of the skin sensitization potential of cosmetics ingredients

Ethical issues in animal toxicity testing have led to the search for alternative methods to determine the skin sensitization potential of cosmetic products. The emergence of ethical testing issues has led to the development of many alternative methods that can reliably estimate skin sensitization potentials. However, a single alternative method may not be able to achieve high predictivity due to the complexity of the skin sensitization mechanism. Therefore, several prediction assays, including both in chemico and in vitro test methods, were investigated and integrated based on the skin sensitization adverse outcome pathway. In this study, we evaluated three different integrated approaches to predict a human skin sensitization hazard using data from in vitro assays (KeratinoSens™ and human cell line activation test [h-CLAT]), and a newly developed in chemico assay (spectrophotometric direct peptide reactivity assay [Spectro-DPRA]). When the results of the in chemico and in vitro assays were combined, the predictivity of human data increased compared with that of a single assay. The highest predictivity was obtained for the approach in which sensitization potential was determined by Spectro-DPRA followed by final determination using the result of KeratinoSens™ and h-CLAT assays (96.3% sensitivity, 87.1% specificity, 86.7% positive predictive value, 96.4% negative predictive value and 91.4% accuracy compared with human data). While further optimization is needed, we believe this integrated approach may provide useful predictive data when determining the human skin sensitization potential of chemicals.

In Silico Approaches in Predictive Genetic Toxicology

Genetic toxicology testing is a weight-of-evidence approach to identify and characterize chemical substances that can cause genetic modifications in somatic and/or germ cells. Prediction of genetic toxicology using computational tools is gaining more attention and preferred by regulatory authorities as an alternate safety assessment for in vivo or in vitro approaches. Due to the cost and time associated with experimental genetic toxicity tests, it is essential to develop more robust in silico methods to predict chemical genetic toxicity. A number of in silico genotoxicity predictive tools/models are developed based on the experimental data gathered over the years. These in silico tools are divided into statistical quantitative structure-activity relationships (QSAR)-based approaches and expert-based systems. This chapter covers the state of the art in silico toxicology approaches and standardized protocols, essential for conducting genetic toxicity predictions of chemicals. This chapter also highlights various parameters for the validation of the prediction results obtained from QSAR models.

The activity of methacrylate esters in skin sensitisation test methods: A review

Skin sensitisation associated with allergic contact dermatitis is an important occupational and environmental disease. The identification of skin sensitisation hazards was traditionally performed using animal tests; originally guinea pig assays and subsequently the murine local lymph node assay (LLNA). More recently there has, for a variety of reasons, been an increased interest in, and requirement for, non-animal assays. There are now available both validated in vitro assays and a variety of approaches based on consideration of quantitative structure-activity relationships (QSAR). With the increased availability and use of non-animal alternatives for skin sensitisation testing there is a continuing need to monitor the performance of these approaches using series of chemicals that do not normally form part of validation exercises. Here we report studies conducted with 11 methacrylate esters and methacrylic acid in which results obtained with 3 validated in vitro tests for which there are OECD guidelines (the Direct Peptide Reactivity Assay, DPRA; ARE-Nrf2 luciferase test methods, and - with some chemicals - a dendritic cell activation test, the myeloid U937 Skin Sensitisation test [U-SENS] assay) have been compared with QSAR approaches (DEREK and TIMES-SS), and with LLNA and guinea pig maximisation test (GPMT) data. The conclusions drawn from these data are that - with this series of chemicals at least - there is a strong correlation between the results of animal tests and the in vitro assays considered, but not with either DEREK or TIMES-SS.

The underlying factors that explain why nucleophilic reagents rarely co-elute with test chemicals in the ADRA

The Amino acid Derivative Reactivity Assay (ADRA) is an in chemico alternative to animal testing for skin sensitization potential that uses two different nucleophilic reagents and it is known that ADRA hardly exhibts co-elution compared with the Direct Peptide Reactivity Assay (DPRA) based on the same scientific principles. In this study, we have analyzed the factors underlying why co-elution, which is sometimes an issue during DPRA testing, virtually never occurs during ADRA testing. Chloramine T and dimethyl isophthalate both exhibited co-elution during DPRA testing, but when quantified at both DPRA's 220 nm and ADRA's 281 nm, we found that when the later detection wavelength was used, these test chemicals produced extremely small peaks that did not interfere with quantification of the peptides. And although both salicylic acid and penicillin G exhibited co-elution during DPRA testing, when tested at a concentration just 1% of that used in DPRA, the very broad peak produced at the higher concentration was reduced significantly. However, both these test chemicals exhibited very sharp peaks when the pH of the injection sample was adjusted to be acidic. Based on these results, we were able to clarify that the reasons why nucleophlic reagents hardly co-elute with test chemicals during ADRA testing are depend on the following three major reasons: (1)differences in the detection wavelength, (2)differences in test chemical concentrations in the injection sample, (3)differences in composition of the injection solvent.

A defined approach for predicting skin sensitisation hazard and potency based on the guided integration of in silico, in chemico and in vitro data using exclusion criteria

A decision tree-based defined approach (DA) has been designed using exclusion criteria based on applicability domain knowledge of in chemico/in vitro information sources covering key events 1-3 in the skin sensitisation adverse outcome pathway and an in silico tool predicting the adverse outcome (Derek Nexus). The hypothesis is that using exclusion criteria to de-prioritise less applicable assays and/or in silico outcomes produces a rational, transparent, and reliable DA for the prediction of skin sensitisation potential. Five exclusion criteria have been established: Derek Nexus reasoning level, Derek Nexus negative prediction, metabolism, lipophilicity, and lysine-reactivity. These are used to prioritise the most suitable information sources for a given chemical and results from which are used in a '2 out of 3' approach to provide a prediction of hazard. A potency category (and corresponding GHS classification) is then assigned using a k-Nearest Neighbours model containing human and LLNA data. The DA correctly identified the hazard (sensitiser/non-sensitiser) for 85% and 86% of a dataset with reference LLNA and human data. The correct potency category was identified for 59% and 68% of chemicals, and the GHS classification accurately predicted for 73% and 76% with reference LLNA and human data, respectively.

Skin and respiratory chemical allergy: confluence and divergence in a hybrid adverse outcome pathway

Sensitisation of the respiratory tract to chemicals resulting in respiratory allergy and allergic asthma is an important occupational health problem, and presents toxicologists with no shortage of challenges. A major issue is that there are no validated or, even widely recognised, methods available for the identification and characterisation of chemical respiratory allergens, or for distinguishing respiratory allergens from contact allergens. The first objective here has been review what is known (and what is not known) of the mechanisms through which chemicals induce sensitisation of the respiratory tract, and to use this information to construct a hybrid Adverse Outcome Pathway (AOP) that combines consideration of both skin and respiratory sensitisation. The intention then has been to use the construction of this hybrid AOP to identify areas of commonality/confluence, and areas of departure/divergence, between skin sensitisation and sensitisation of the respiratory tract. The hybrid AOP not only provides a mechanistic understanding of how the processes of skin and respiratory sensitisation differ, buy also a means of identifying areas of uncertainty about chemical respiratory allergy that benefit from a further investment in research.

Non-animal methods to predict skin sensitization (II): an assessment of defined approaches *

Skin sensitization is a toxicity endpoint of widespread concern, for which the mechanistic understanding and concurrent necessity for non-animal testing approaches have evolved to a critical juncture, with many available options for predicting sensitization without using animals. Cosmetics Europe and the National Toxicology Program Interagency Center for the Evaluation of Alternative Toxicological Methods collaborated to analyze the performance of multiple non-animal data integration approaches for the skin sensitization safety assessment of cosmetics ingredients. The Cosmetics Europe Skin Tolerance Task Force (STTF) collected and generated data on 128 substances in multiple in vitro and in chemico skin sensitization assays selected based on a systematic assessment by the STTF. These assays, together with certain in silico predictions, are key components of various non-animal testing strategies that have been submitted to the Organization for Economic Cooperation and Development as case studies for skin sensitization. Curated murine local lymph node assay (LLNA) and human skin sensitization data were used to evaluate the performance of six defined approaches, comprising eight non-animal testing strategies, for both hazard and potency characterization. Defined approaches examined included consensus methods, artificial neural networks, support vector machine models, Bayesian networks, and decision trees, most of which were reproduced using open source software tools. Multiple non-animal testing strategies incorporating in vitro, in chemico, and in silico inputs demonstrated equivalent or superior performance to the LLNA when compared to both animal and human data for skin sensitization.

Tools to investigate and avoid drug-hypersensitivity in drug development

INTRODUCTION

Drug hypersensitivity reactions (DHRs) are common adverse effects of pharmaceuticals that clinically resemble allergies, and which are becoming an important burden to healthcare systems. Alongside accurate diagnostic techniques, tools which can predict potential drug-inducing hypersensitivity reactions in the pre-clinical phase are critical. Despite the important adverse reactions linked to immune-mediated hypersensitivity, at present, there are no validated or required in vivo or in vitro methods to screen the sensitizing potential of drugs and their metabolites in the pre-clinical phase. Areas covered: Enhanced prediction in preclinical safety evaluation is extremely important. The purpose of this review is to assess the state of the art of tools available to assess the allergenic potential of drugs and to highlight our current understanding of the molecular mechanisms underlying inappropriate immune activation. Expert opinion: The knowledge that allergenic drugs share common mechanisms of immune cell activation with chemical allergens, and of the definition of the mechanistic pathway to adverse outcomes, can enhance targeting toxicity testing in drug development and hazard assessment of hypersensitivity. Additional efforts and extensive resources are necessary to improve preclinical testing methodologies, including optimization, better design and interpretation of data.

Behaviour of chemical respiratory allergens in novel predictive methods for skin sensitisation

Asthma resulting from sensitisation of the respiratory tract to chemicals is an important occupational health issue, presenting many toxicological challenges. Most importantly there are no recognised predictive methods for respiratory allergens. Nevertheless, it has been found that all known chemical respiratory allergens elicit positive responses in assays for skin sensitising chemicals. Thus, chemicals failing to induce a positive response in skin sensitisation assays such as the local lymph node assay (LLNA) lack not only skin sensitising activity, but also the potential to cause respiratory sensitisation. However, it is unclear whether it will be possible to regard chemicals that are negative in in vitro skin sensitisation tests also as lacking respiratory sensitising activity. To address this, the behaviour of chemical respiratory allergens in the LLNA and in recently validated non-animal tests for skin sensitisation have been examined. Most chemical respiratory allergens are positive in one or more newly validated non-animal test methods, although the situation varies between individual assays. The use of an integrated testing strategy could provide a basis for recognition of most respiratory sensitising chemicals. However, a more complete picture of the performance characteristics of such tests is required before specific recommendations can be made.

The adverse outcome pathway for skin sensitisation: Moving closer to replacing animal testing

This article outlines the work of the Organisation for Economic Co-operation and Development (OECD) that led to being jointly awarded the 2015 Lush Black Box Prize. The award-winning work centred on the development of 'The Adverse Outcome Pathway for Skin Sensitisation Initiated by Covalent Binding to Proteins'. This Adverse Outcome Pathway (AOP) has provided the mechanistic basis for the integration of skin sensitisation-related information. Recent developments in integrated approaches to testing and assessment, based on the AOP, are summarised. The impact of the AOP on regulatory policy and on the Three Rs are discussed. An overview of the next generation of the skin sensitisation AOP module in the OECD QSAR Toolbox, based on more-recent work at the Laboratory of Mathematical Chemistry, is also presented.

Insights on in vitro models for safety and toxicity assessment of cosmetic ingredients

According to the current European legislation, the safety assessment of each individual cosmetic ingredient of any formulation is the basis for the safety evaluation of a cosmetic product. Also, animal testing in the European Union is prohibited for cosmetic ingredients and products since 2004 and 2009, respectively. Additionally, the commercialization of any cosmetic products containing ingredients tested on animal models was forbidden in 2009. In consequence of these boundaries, the European Centre for the Validation of Alternative Methods (ECVAM) proposes a list of validated cell-based in vitro models for predicting the safety and toxicity of cosmetic ingredients. These models have been demonstrated as valuable and effective tools to overcome the limitations of animal in vivo studies. Although the use of in vitro cell-based models for the evaluation of absorption and permeability of cosmetic ingredients is widespread, a detailed study on the properties of these platforms and the in vitro-in vivo correlation compared with human data are required. Moreover, additional efforts must be taken to develop in vitro models to predict carcinogenicity, repeat dose toxicity and reproductive toxicity, for which no alternative in vitro methods are currently available. This review paper summarizes and characterizes the most relevant in vitro models validated by ECVAM employed to predict the safety and toxicology of cosmetic ingredients.

Evaluation of in silico tools to predict the skin sensitization potential of chemicals

Public domain and commercial in silico tools were compared for their performance in predicting the skin sensitization potential of chemicals. The packages were either statistical based (Vega, CASE Ultra) or rule based (OECD Toolbox, Toxtree, Derek Nexus). In practice, several of these in silico tools are used in gap filling and read-across, but here their use was limited to make predictions based on presence/absence of structural features associated to sensitization. The top 400 ranking substances of the ATSDR 2011 Priority List of Hazardous Substances were selected as a starting point. Experimental information was identified for 160 chemically diverse substances (82 positive and 78 negative). The prediction for skin sensitization potential was compared with the experimental data. Rule-based tools perform slightly better, with accuracies ranging from 0.6 (OECD Toolbox) to 0.78 (Derek Nexus), compared with statistical tools that had accuracies ranging from 0.48 (Vega) to 0.73 (CASE Ultra - LLNA weak model). Combining models increased the performance, with positive and negative predictive values up to 80% and 84%, respectively. However, the number of substances that were predicted positive or negative for skin sensitization in both models was low. Adding more substances to the dataset will increase the confidence in the conclusions reached. The insights obtained in this evaluation are incorporated in a web database www.asopus.weebly.com that provides a potential end user context for the scope and performance of different in silico tools with respect to a common dataset of curated skin sensitization data.

State of the art in non-animal approaches for skin sensitization testing: from individual test methods towards testing strategies

The hazard assessment of skin sensitizers relies mainly on animal testing, but much progress is made in the development, validation and regulatory acceptance and implementation of non-animal predictive approaches. In this review, we provide an update on the available computational tools and animal-free test methods for the prediction of skin sensitization hazard. These individual test methods address mostly one mechanistic step of the process of skin sensitization induction. The adverse outcome pathway (AOP) for skin sensitization describes the key events (KEs) that lead to skin sensitization. In our review, we have clustered the available test methods according to the KE they inform: the molecular initiating event (MIE/KE1)-protein binding, KE2-keratinocyte activation, KE3-dendritic cell activation and KE4-T cell activation and proliferation. In recent years, most progress has been made in the development and validation of in vitro assays that address KE2 and KE3. No standardized in vitro assays for T cell activation are available; thus, KE4 cannot be measured in vitro. Three non-animal test methods, addressing either the MIE, KE2 or KE3, are accepted as OECD test guidelines, and this has accelerated the development of integrated or defined approaches for testing and assessment (e.g. testing strategies). The majority of these approaches are mechanism-based, since they combine results from multiple test methods and/or computational tools that address different KEs of the AOP to estimate skin sensitization potential and sometimes potency. Other approaches are based on statistical tools. Until now, eleven different testing strategies have been published, the majority using the same individual information sources. Our review shows that some of the defined approaches to testing and assessment are able to accurately predict skin sensitization hazard, sometimes even more accurate than the currently used animal test. A few defined approaches are developed to provide an estimate of the potency sub-category of a skin sensitizer as well, but these approaches need further independent evaluation with a new dataset of chemicals. To conclude, this update shows that the field of non-animal approaches for skin sensitization has evolved greatly in recent years and that it is possible to predict skin sensitization hazard without animal testing.