Bayesian integrated testing strategy to assess skin sensitization potency: from theory to practice

Deriving a point of departure for assessing the skin sensitization risk of wearable device constituents with in vitro methods

Wearable devices are in contact with the skin for extended periods. As such, the device constituents should be evaluated for their skin sensitization potential, and a Point of Departure (PoD) should be derived to conduct a proper risk assessment. Without historical in vivo data, the PoD must be derived with New Approach Methods (NAMs). To accomplish this, regression models trained on LLNA data that use data inputs from OECD-validated in vitro tests were used to derive a predicted EC3 value, the LLNA value used to classify skin sensitization potency, for three adhesive monomers (Isobornyl acrylate (IBOA), N, N- Dimethylacrylamide (NNDMA), and Acryloylmorpholine (ACMO) and one dye (Solvent Orange 60 (SO60)). These chemicals can be used as constituents of wearable devices and have been associated with causing allergic contact dermatitis (ACD). Using kinetic DPRA and KeratinoSens™ data, the PoDs obtained with the regression model were 180, 215, 1535, and 8325 μg/cm2 for IBOA, SO60, ACMO, and NNDMA, respectively. The PoDs derived with the regression model using NAMs data will enable a proper skin sensitization risk assessment without using animals.

Protein Haptenation and Its Role in Allergy

Humans are exposed to numerous electrophilic chemicals either as medicines, in the workplace, in nature, or through use of many common cosmetic and household products. Covalent modification of human proteins by such chemicals, or protein haptenation, is a common occurrence in cells and may result in generation of antigenic species, leading to development of hypersensitivity reactions. Ranging in severity of symptoms from local cutaneous reactions and rhinitis to potentially life-threatening anaphylaxis and severe hypersensitivity reactions such as Stephen-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN), all these reactions have the same Molecular Initiating Event (MIE), i.e. haptenation. However, not all individuals who are exposed to electrophilic chemicals develop symptoms of hypersensitivity. In the present review, we examine common chemistry behind the haptenation reactions leading to formation of neoantigens. We explore simple reactions involving single molecule additions to a nucleophilic side chain of proteins and complex reactions involving multiple electrophilic centers on a single molecule or involving more than one electrophilic molecule as well as the generation of reactive molecules from the interaction with cellular detoxification mechanisms. Besides generation of antigenic species and enabling activation of the immune system, we explore additional events which result directly from the presence of electrophilic chemicals in cells, including activation of key defense mechanisms and immediate consequences of those reactions, and explore their potential effects. We discuss the factors that work in concert with haptenation leading to the development of hypersensitivity reactions and those that may act to prevent it from developing. We also review the potential harnessing of the specificity of haptenation in the design of potent covalent therapeutic inhibitors.

Bayesian analysis suggests independent development of sensitization to different fungal allergens

Background

Fungi are known for their ability to cause allergies, but data on individual sensitization to them are insufficient. The purpose of the study was to carry out a comprehensive analysis of the fungal allergens' sensitization profile in the Ukrainian population and to determine both population and individual sensitivity to these allergens.

Methods

We utilized a set of ALEX allergy test data from 20,033 inhabitants of 17 regions of Ukraine from 1 to 89 years conducted in 2020-2022. A complex of programs in the Python language was developed and Bayesian network analysis was applied to determine the sensitivity combinations in individual patients to various fungal components.

Results

Sensitivity to Alt a 1 dominated and was observed in 79.39% of patients, and 62.17% of them were sensitive solely to Alt a 1. Exclusive sensitivity to Mala s 6 was second in individual patient profiles with a frequency of 4.06%. Combined sensitivity to Alt a 1 - Asp f 3 was third with a share of 3.28%. Pen ch and Cla h extracts stimulated the production of the lowest median sIgE levels. The highest median sIgE levels were for Alt a 1, Mala s 11 and Asp f 6, respectively. Median sIgE levels increased in adults compared to children for all components of Aspergillus fumigatus, as well as for Mala s 5 and Mala s 11. In the rest of the cases, they decreased in adults compared to children. The sensitization rates to fungi in general and specifically to Alternaria were lower in the western parts of Ukraine, especially in the Carpathian region, situated within the Broad-leaved Forest zone. The results of Bayesian modeling revealed that in the case of Alt a 1, the simultaneous absence of sensitivity to Cla h 8, Mala s 11, Mala s 5 and Mala s 6 molecules could condition the presence of sensitization to the major Alternaria allergen with a probability of 92.42%. In all other cases, there was a high probability of absence of sensitivity to particular allergen against the background of absence of sensitivity to other ones, which may indicate the independent development of sensitization to different fungal allergens.

Conclusions

Sensitivity to Alt a 1 dominated in the studied population with a lower rate in the western regions. The highest median sIgE levels were induced by Alt a 1, Mala s 11 and Asp f 6. Bayesian Analysis suggest a high probability of the independent development of sensitization to different fungal allergens. The idea that sensitization to one allergen may be protective against sensitization to another one(s) requires further clinical study.

The skin sensitization adverse outcome pathway: exploring the role of mechanistic understanding for higher tier risk assessment

For over a decade, the skin sensitization Adverse Outcome Pathway (AOP) has served as a useful framework for development of novel in chemico and in vitro assays for use in skin sensitization hazard and risk assessment. Since its establishment, the AOP framework further fueled the existing efforts in new assay development and stimulated a plethora of activities with particular focus on validation, reproducibility and interpretation of individual assays and combination of assay outputs for use in hazard/risk assessment. In parallel, research efforts have also accelerated in pace, providing new molecular and dynamic insight into key events leading to sensitization. In light of novel hypotheses emerging from over a decade of focused research effort, mechanistic evidence relating to the key events in the skin sensitization AOP may complement the tools currently used in risk assessment. We reviewed the recent advances unraveling the complexity of molecular events in sensitization and signpost the most promising avenues for further exploration and development of useful assays.

Sensitization patterns to Poaceae pollen indicates a hierarchy in allergens and a lead of tropical grasses

BACKGROUND

The allergenicity of pollen of Poaceae family members is a well-known and confirmed fact. Using the data of component-resolved molecular diagnostics of allergy, we set a goal to establish the population and individual characteristics of sensitization to grass pollen and assess the patterns of its development.

METHODS

Multiplex allergy Alex2 test results of 20,033 patients were used. In addition to descriptive statistics to uncover traits of the sensitized population, statistical inference was utilized to establish the conditional probability of sensitisation, the nature of links between allergens, and the most frequent combinations of allergens in individual patient profiles.

RESULTS

Sensitivity to grass pollen comprised 30.79% of the studied sample. Children accounted for 62.21%, adults-37.79%. Sensitisation to Phl p 1, Lol p 1, and Cyn d 1 was the most frequent in all age groups. Among them, Phl p 1 and Lol p 1 were the major ones. Phl p 2, Phl p 5.0101, and Phl p 6 were also responsible for primary sensitization; Phl p 5.0101 promoted the highest sIgE levels. A combination "Lol p 1-Phl p 1", where Lol p 1 might play a leading role, was most frequent in individual profiles. Monosensitization to Phl p 2 was the second most frequent and Bayesian Network suggested its independent development. Monosensitization to Cyn d 1, especially among children, may indicate the impact of climate change, promoting the spread of the subtropical grasses to the temperate region.

CONCLUSIONS

Descriptive statistics and known clinical data coincide well with statistical inference results and can provide for new clinical insights.

Application of direct peptide reactivity assay for assessing the skin sensitization potential of essential oils

Plant-derived products are frequently found as ingredients in cosmetics. However, the current data show non-neglectable skin sensitizing potential of these preparations suggesting an urgent need for data regarding their health safety profile. The aim of this study was to assess the skin sensitization potential of commercial essential oils by selected Lamiaceae species (Lavandula angustifolia, Melissa officinalis, Mentha longifolia, Thymus vulgaris, Salvia officinalis, and Rosmarinus officinalis) using a chemistry-based Direct Peptide Reactivity Assay (DPRA) in order to predict their potential allergic properties. In the DPRA assay, nucleophile-containing synthetic peptides (cysteine peptide and lysine peptide) were incubated with the test substance for 24 h. Depletion of the peptide in the reaction mixture was measured by high-pressure liquid chromatography (HPLC) using UV detection and the average peptide depletion data for cysteine and lysine was then calculated. Menthae longifoliae aetheroleum showed no or minimal reactivity with 4.48% cysteine depletion, Rosmarini aetheroleum and Salviae aetheroleum showed low reactivity with the 12.79% and 15.34% of cysteine depletion, respectively, while the other analyzed essential oils showed moderate reactivity with the cysteine depletion between 23.21 and 48.43%. According to DPRA predictive analysis, only Menthae longifoliae aetheroleum can be classified as negative, while all other essential oils may be classified as positive, thus having the potential to cause skin sensitization.

Mapping Mechanistic Pathways of Acute Oral Systemic Toxicity Using Chemical Structure and Bioactivity Measurements

Regulatory agencies around the world have committed to reducing or eliminating animal testing for establishing chemical safety. Adverse outcome pathways can facilitate replacement by providing a mechanistic framework for identifying the appropriate non-animal methods and connecting them to apical adverse outcomes. This study separated 11,992 chemicals with curated rat oral acute toxicity information into clusters of structurally similar compounds. Each cluster was then assigned one or more ToxCast/Tox21 assays by looking for the minimum number of assays required to record at least one positive hit call below cytotoxicity for all acutely toxic chemicals in the cluster. When structural information is used to select assays for testing, none of the chemicals required more than four assays and 98% required two assays or less. Both the structure-based clusters and activity from the associated assays were significantly associated with the GHS toxicity classification of the chemicals, which suggests that a combination of bioactivity and structural information could be as reproducible as traditional in vivo studies. Predictivity is improved when the in vitro assay directly corresponds to the mechanism of toxicity, but many indirect assays showed promise as well. Given the lower cost of in vitro testing, a small assay battery including both general cytotoxicity assays and two or more orthogonal assays targeting the toxicological mechanism could be used to improve performance further. This approach illustrates the promise of combining existing in silico approaches, such as the Collaborative Acute Toxicity Modeling Suite (CATMoS), with structure-based bioactivity information as part of an efficient tiered testing strategy that can reduce or eliminate animal testing for acute oral toxicity.

A framework for chemical safety assessment incorporating new approach methodologies within REACH

The long-term investment in new approach methodologies (NAMs) within the EU and other parts of the world is beginning to result in an emerging consensus of how to use information from in silico, in vitro and targeted in vivo sources to assess the safety of chemicals. However, this methodology is being adopted very slowly for regulatory purposes. Here, we have developed a framework incorporating in silico, in vitro and in vivo methods designed to meet the requirements of REACH in which both hazard and exposure can be assessed using a tiered approach. The outputs from each tier are classification categories, safe doses, and risk assessments, and progress through the tiers depends on the output from previous tiers. We have exemplified the use of the framework with three examples. The outputs were the same or more conservative than parallel assessments based on conventional studies. The framework allows a transparent and phased introduction of NAMs in chemical safety assessment and enables science-based safety decisions which provide the same level of public health protection using fewer animals, taking less time, and using less financial and expert resource. Furthermore, it would also allow new methods to be incorporated as they develop through continuous selective evolution rather than periodic revolution.

Probabilistic risk assessment - the keystone for the future of toxicology

Safety sciences must cope with uncertainty of models and results as well as information gaps. Acknowledging this uncertainty necessitates embracing probabilities and accepting the remaining risk. Every toxicological tool delivers only probable results. Traditionally, this is taken into account by using uncertainty / assessment factors and worst-case / precautionary approaches and thresholds. Probabilistic methods and Bayesian approaches seek to characterize these uncertainties and promise to support better risk assessment and, thereby, improve risk management decisions. Actual assessments of uncertainty can be more realistic than worst-case scenarios and may allow less conservative safety margins. Most importantly, as soon as we agree on uncertainty, this defines room for improvement and allows a transition from traditional to new approach methods as an engineering exercise. The objective nature of these mathematical tools allows to assign each methodology its fair place in evidence integration, whether in the context of risk assessment, systematic reviews, or in the definition of an integrated testing strategy (ITS) / defined approach (DA) / integrated approach to testing and assessment (IATA). This article gives an overview of methods for probabilistic risk assessment and their application for exposure assessment, physiologically-based kinetic modelling, probability of hazard assessment (based on quantitative and read-across based structure-activity relationships, and mechanistic alerts from in vitro studies), individual susceptibility assessment, and evidence integration. Additional aspects are opportunities for uncertainty analysis of adverse outcome pathways and their relation to thresholds of toxicological concern. In conclusion, probabilistic risk assessment will be key for constructing a new toxicology paradigm – probably!

In silico Prediction of Skin Sensitization: Quo vadis?

Skin direct contact with chemical or physical substances is predisposed to allergic contact dermatitis (ACD), producing various allergic reactions, namely rash, blister, or itchy, in the contacted skin area. ACD can be triggered by various extremely complicated adverse outcome pathways (AOPs) remains to be causal for biosafety warrant. As such, commercial products such as ointments or cosmetics can fulfill the topically safe requirements in animal and non-animal models including allergy. Europe, nevertheless, has banned animal tests for the safety evaluations of cosmetic ingredients since 2013, followed by other countries. A variety of non-animal in vitro tests addressing different key events of the AOP, the direct peptide reactivity assay (DPRA), KeratinoSens™, LuSens and human cell line activation test h-CLAT and U-SENS™ have been developed and were adopted in OECD test guideline to identify the skin sensitizers. Other methods, such as the SENS-IS are not yet fully validated and regulatorily accepted. A broad spectrum of in silico models, alternatively, to predict skin sensitization have emerged based on various animal and non-animal data using assorted modeling schemes. In this article, we extensively summarize a number of skin sensitization predictive models that can be used in the biopharmaceutics and cosmeceuticals industries as well as their future perspectives, and the underlined challenges are also 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.

Building and Applying Quantitative Adverse Outcome Pathway Models for Chemical Hazard and Risk Assessment

An important goal in toxicology is the development of new ways to increase the speed, accuracy, and applicability of chemical hazard and risk assessment approaches. A promising route is the integration of in vitro assays with biological pathway information. We examined how the adverse outcome pathway (AOP) framework can be used to develop pathway-based quantitative models useful for regulatory chemical safety assessment. By using AOPs as initial conceptual models and the AOP knowledge base as a source of data on key event relationships, different methods can be applied to develop computational quantitative AOP models (qAOPs) relevant for decision making. A qAOP model may not necessarily have the same structure as the AOP it is based on. Useful AOP modeling methods range from statistical, Bayesian networks, regression, and ordinary differential equations to individual-based models and should be chosen according to the questions being asked and the data available. We discuss the need for toxicokinetic models to provide linkages between exposure and qAOPs, to extrapolate from in vitro to in vivo, and to extrapolate across species. Finally, we identify best practices for modeling and model building and the necessity for transparent and comprehensive documentation to gain confidence in the use of qAOP models and ultimately their use in regulatory applications. Environ Toxicol Chem 2019;38:1850-1865. © 2019 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals, Inc. on behalf of SETAC.

Big-data and machine learning to revamp computational toxicology and its use in risk assessment

The creation of large toxicological databases and advances in machine-learning techniques have empowered computational approaches in toxicology. Work with these large databases based on regulatory data has allowed reproducibility assessment of animal models, which highlight weaknesses in traditional in vivo methods. This should lower the bars for the introduction of new approaches and represents a benchmark that is achievable for any alternative method validated against these methods. Quantitative Structure Activity Relationships (QSAR) models for skin sensitization, eye irritation, and other human health hazards based on these big databases, however, also have made apparent some of the challenges facing computational modeling, including validation challenges, model interpretation issues, and model selection issues. A first implementation of machine learning-based predictions termed REACHacross achieved unprecedented sensitivities of >80% with specificities >70% in predicting the six most common acute and topical hazards covering about two thirds of the chemical universe. While this is awaiting formal validation, it demonstrates the new quality introduced by big data and modern data-mining technologies. The rapid increase in the diversity and number of computational models, as well as the data they are based on, create challenges and opportunities for the use of computational methods.

The Adverse Outcome Pathway: A Multifaceted Framework Supporting 21st Century Toxicology

The adverse outcome pathway (AOP) framework serves as a knowledge assembly, interpretation, and communication tool designed to support the translation of pathway-specific mechanistic data into responses relevant to assessing and managing risks of chemicals to human health and the environment. As such, AOPs facilitate the use of data streams often not employed by risk assessors, including information from in silico models, in vitro assays and short-term in vivo tests with molecular/biochemical endpoints. This translational capability can increase the capacity and efficiency of safety assessments both for single chemicals and chemical mixtures. Our mini-review describes the conceptual basis of the AOP framework and aspects of its current status relative to use by toxicologists and risk assessors, including four illustrative applications of the framework to diverse assessment scenarios.

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.

Determination of Protein Haptenation by Chemical Sensitizers Within the Complexity of the Human Skin Proteome

Skin sensitization associated with the development of allergic contact dermatitis occurs via a number of specific key events at the cellular level. The molecular initiating event (MIE), the first in the sequence of these events, occurs after exposure of the skin to an electrophilic chemical, causing the irreversible haptenation of proteins within skin. Characterization of this MIE is a key step in elucidating the skin sensitization adverse outcome pathway and is essential to providing parameters for mathematical models to predict the capacity of a chemical to cause sensitization. As a first step to addressing this challenge, we have exposed complex protein lysates from a keratinocyte cell line and human skin tissue with a range of well characterized sensitizers, including dinitrochlorobenzene, 5-chloro-2-methylisothiazol-3-one, cinnamaldehyde, and the non (or weak) sensitizer 6-methyl coumarin. Using a novel stable isotope labeling approach combined with ion mobility-assisted data independent mass spectrometry (HDMSE), we have characterized the haptenome for these sensitizers. Although a significant proportion of highly abundant proteins were haptenated, we also observed the haptenation of low abundant proteins by all 3 of the chemical sensitizers tested, indicating that within a complex protein background, protein abundance is not the sole determinant driving haptenation, highlighting a relationship to tertiary protein structure and the amino acid specificity of these chemical sensitizers and sensitizer potency.

Comparison and validation of an in vitro skin sensitization strategy using a data set of 33 chemical references

Allergic contact dermatitis (ACD) is an adverse health effect that develops following repeated exposure to skin sensitizing chemicals. An animal testing ban has been applied in EU, leading to development of reliably predictive non-animal methods. Several in vitro methods have been developed as alternatives but one single non-animal test method is not been sufficient to fully address since the LLNA test ban. Here, we have selected an ITS (Integrated Testing Strategy) for skin sensitization which focuses on three in vitro methods that covered the first three steps of the AOP (DPRA, SENS-IS or h-CLAT). The aim of this study was to compare these three methods due to the WoE approach based on a 2-out-of-3-assessment. The results of 33 references were compared to in vivo data (especially human). We have shown that tested firstly DPRA and SENS-IS have permitted to conclude on 29 of 33 chemicals, whereas DPRA and h-CLAT on 25, and SENS-IS and h-CLAT on 23. With this sequence, DPRA and SENS-IS and then h-CLAT in case of equivocal results, we conclude more quickly by performing fewer tests. Thereby, we have shown that it is better to follow a preferential sequence than testing chemicals simultaneously with these three methods.

How Adverse Outcome Pathways Can Aid the Development and Use of Computational Prediction Models for Regulatory Toxicology

Efforts are underway to transform regulatory toxicology and chemical safety assessment from a largely empirical science based on direct observation of apical toxicity outcomes in whole organism toxicity tests to a predictive one in which outcomes and risk are inferred from accumulated mechanistic understanding. The adverse outcome pathway (AOP) framework provides a systematic approach for organizing knowledge that may support such inference. Likewise, computational models of biological systems at various scales provide another means and platform to integrate current biological understanding to facilitate inference and extrapolation. We argue that the systematic organization of knowledge into AOP frameworks can inform and help direct the design and development of computational prediction models that can further enhance the utility of mechanistic and in silico data for chemical safety assessment. This concept was explored as part of a workshop on AOP-Informed Predictive Modeling Approaches for Regulatory Toxicology held September 24-25, 2015. Examples of AOP-informed model development and its application to the assessment of chemicals for skin sensitization and multiple modes of endocrine disruption are provided. The role of problem formulation, not only as a critical phase of risk assessment, but also as guide for both AOP and complementary model development is described. Finally, a proposal for actively engaging the modeling community in AOP-informed computational model development is made. The contents serve as a vision for how AOPs can be leveraged to facilitate development of computational prediction models needed to support the next generation of chemical safety assessment.

Testing Human Skin and Respiratory Sensitizers-What Is Good Enough?

Alternative methods for accurate in vitro assessment of skin and respiratory sensitizers are urgently needed. Sensitization is a complex biological process that cannot be evaluated accurately using single events or biomarkers, since the information content is too restricted in these measurements. On the contrary, if the tremendous information content harbored in DNA/mRNA could be mined, most complex biological processes could be elucidated. Genomic technologies available today, including transcriptional profiling and next generation sequencing, have the power to decipher sensitization, when used in the right context. Thus, a genomic test platform has been developed, denoted the Genomic Allergen Rapid Detection (GARD) assay. Due to the high informational content of the GARD test, accurate predictions of both the skin and respiratory sensitizing capacity of chemicals, have been demonstrated. Based on a matured dendritic cell line, acting as a human-like reporter system, information about potency has also been acquired. Consequently, multiparametric diagnostic technologies are disruptive test principles that can change the way in which the next generation of alternative methods are designed.

Prediction of skin sensitization potency using machine learning approaches

The replacement of animal use in testing for regulatory classification of skin sensitizers is a priority for US federal agencies that use data from such testing. Machine learning models that classify substances as sensitizers or non-sensitizers without using animal data have been developed and evaluated. Because some regulatory agencies require that sensitizers be further classified into potency categories, we developed statistical models to predict skin sensitization potency for murine local lymph node assay (LLNA) and human outcomes. Input variables for our models included six physicochemical properties and data from three non-animal test methods: direct peptide reactivity assay; human cell line activation test; and KeratinoSens™ assay. Models were built to predict three potency categories using four machine learning approaches and were validated using external test sets and leave-one-out cross-validation. A one-tiered strategy modeled all three categories of response together while a two-tiered strategy modeled sensitizer/non-sensitizer responses and then classified the sensitizers as strong or weak sensitizers. The two-tiered model using the support vector machine with all assay and physicochemical data inputs provided the best performance, yielding accuracy of 88% for prediction of LLNA outcomes (120 substances) and 81% for prediction of human test outcomes (87 substances). The best one-tiered model predicted LLNA outcomes with 78% accuracy and human outcomes with 75% accuracy. By comparison, the LLNA predicts human potency categories with 69% accuracy (60 of 87 substances correctly categorized). These results suggest that computational models using non-animal methods may provide valuable information for assessing skin sensitization potency. Copyright © 2017 John Wiley & Sons, Ltd.

Integrated decision strategies for skin sensitization hazard

One of the top priorities of the Interagency Coordinating Committee for the Validation of Alternative Methods (ICCVAM) is the identification and evaluation of non-animal alternatives for skin sensitization testing. Although skin sensitization is a complex process, the key biological events of the process have been well characterized in an adverse outcome pathway (AOP) proposed by the Organisation for Economic Co-operation and Development (OECD). Accordingly, ICCVAM is working to develop integrated decision strategies based on the AOP using in vitro, in chemico and in silico information. Data were compiled for 120 substances tested in the murine local lymph node assay (LLNA), direct peptide reactivity assay (DPRA), human cell line activation test (h-CLAT) and KeratinoSens assay. Data for six physicochemical properties, which may affect skin penetration, were also collected, and skin sensitization read-across predictions were performed using OECD QSAR Toolbox. All data were combined into a variety of potential integrated decision strategies to predict LLNA outcomes using a training set of 94 substances and an external test set of 26 substances. Fifty-four models were built using multiple combinations of machine learning approaches and predictor variables. The seven models with the highest accuracy (89-96% for the test set and 96-99% for the training set) for predicting LLNA outcomes used a support vector machine (SVM) approach with different combinations of predictor variables. The performance statistics of the SVM models were higher than any of the non-animal tests alone and higher than simple test battery approaches using these methods. These data suggest that computational approaches are promising tools to effectively integrate data sources to identify potential skin sensitizers without animal testing. Published 2016. This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Multivariate models for prediction of human skin sensitization hazard

One of the Interagency Coordinating Committee on the Validation of Alternative Method's (ICCVAM) top priorities is the development and evaluation of non-animal approaches to identify potential skin sensitizers. The complexity of biological events necessary to produce skin sensitization suggests that no single alternative method will replace the currently accepted animal tests. ICCVAM is evaluating an integrated approach to testing and assessment based on the adverse outcome pathway for skin sensitization that uses machine learning approaches to predict human skin sensitization hazard. We combined data from three in chemico or in vitro assays - the direct peptide reactivity assay (DPRA), human cell line activation test (h-CLAT) and KeratinoSens™ assay - six physicochemical properties and an in silico read-across prediction of skin sensitization hazard into 12 variable groups. The variable groups were evaluated using two machine learning approaches, logistic regression and support vector machine, to predict human skin sensitization hazard. Models were trained on 72 substances and tested on an external set of 24 substances. The six models (three logistic regression and three support vector machine) with the highest accuracy (92%) used: (1) DPRA, h-CLAT and read-across; (2) DPRA, h-CLAT, read-across and KeratinoSens; or (3) DPRA, h-CLAT, read-across, KeratinoSens and log P. The models performed better at predicting human skin sensitization hazard than the murine local lymph node assay (accuracy 88%), any of the alternative methods alone (accuracy 63-79%) or test batteries combining data from the individual methods (accuracy 75%). These results suggest that computational methods are promising tools to identify effectively the potential human skin sensitizers without animal testing. Published 2016. This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Analysis of publically available skin sensitization data from REACH registrations 2008-2014

The public data on skin sensitization from REACH registrations already included 19,111 studies on skin sensitization in December 2014, making it the largest repository of such data so far (1,470 substances with mouse LLNA, 2,787 with GPMT, 762 with both in vivo and in vitro and 139 with only in vitro data). 21% were classified as sensitizers. The extracted skin sensitization data was analyzed to identify relationships in skin sensitization guidelines, visualize structural relationships of sensitizers, and build models to predict sensitization.

A chemical with molecular weight > 500 Da is generally considered non-sensitizing owing to low bioavailability, but 49 sensitizing chemicals with a molecular weight > 500 Da were found.

A chemical similarity map was produced using PubChem’s 2D Tanimoto similarity metric and Gephi force layout visualization. Nine clusters of chemicals were identified by Blondel’s module recognition algorithm revealing wide module-dependent variation.

Approximately 31% of mapped chemicals are Michael’s acceptors but alone this does not imply skin sensitization. A simple sensitization model using molecular weight and five ToxTree structural alerts showed a balanced accuracy of 65.8% (specificity 80.4%, sensitivity 51.4%), demonstrating that structural alerts have information value.

A simple variant of k-nearest neighbors outperformed the ToxTree approach even at 75% similarity threshold (82% balanced accuracy at 0.95 threshold). At higher thresholds, the balanced accuracy increased. Lower similarity thresholds decrease sensitivity faster than specificity.

This analysis scopes the landscape of chemical skin sensitization, demonstrating the value of large public datasets for health hazard prediction.

Probabilistic hazard assessment for skin sensitization potency by dose-response modeling using feature elimination instead of quantitative structure-activity relationships

Supervised learning methods promise to improve integrated testing strategies (ITS), but must be adjusted to handle high dimensionality and dose-response data. ITS approaches are currently fueled by the increasing mechanistic understanding of adverse outcome pathways (AOP) and the development of tests reflecting these mechanisms. Simple approaches to combine skin sensitization data sets, such as weight of evidence, fail due to problems in information redundancy and high dimensionality. The problem is further amplified when potency information (dose/response) of hazards would be estimated. Skin sensitization currently serves as the foster child for AOP and ITS development, as legislative pressures combined with a very good mechanistic understanding of contact dermatitis have led to test development and relatively large high-quality data sets. We curated such a data set and combined a recursive variable selection algorithm to evaluate the information available through in silico, in chemico and in vitro assays. Chemical similarity alone could not cluster chemicals' potency, and in vitro models consistently ranked high in recursive feature elimination. This allows reducing the number of tests included in an ITS. Next, we analyzed with a hidden Markov model that takes advantage of an intrinsic inter-relationship among the local lymph node assay classes, i.e. the monotonous connection between local lymph node assay and dose. The dose-informed random forest/hidden Markov model was superior to the dose-naive random forest model on all data sets. Although balanced accuracy improvement may seem small, this obscures the actual improvement in misclassifications as the dose-informed hidden Markov model strongly reduced " false-negatives" (i.e. extreme sensitizers as non-sensitizer) on all data sets.

Evaluation of combinations of in vitro sensitization test descriptors for the artificial neural network-based risk assessment model of skin sensitization

The skin sensitization potential of chemicals has been determined with the use of the murine local lymph node assay (LLNA). However, in recent years public concern about animal welfare has led to a requirement for non-animal risk assessment systems for the prediction of skin sensitization potential, to replace LLNA. Selection of an appropriate in vitro test or in silico model descriptors is critical to obtain good predictive performance. Here, we investigated the utility of artificial neural network (ANN) prediction models using various combinations of descriptors from several in vitro sensitization tests. The dataset, collected from published data and from experiments carried out in collaboration with the Japan Cosmetic Industry Association (JCIA), consisted of values from the human cell line activation test (h-CLAT), direct peptide reactivity assay (DPRA), SH test and antioxidant response element (ARE) assay for chemicals whose LLNA thresholds have been reported. After confirming the relationship between individual in vitro test descriptors and the LLNA threshold (e.g. EC3 value), we used the subsets of chemicals for which the requisite test values were available to evaluate the predictive performance of ANN models using combinations of h-CLAT/DPRA (N = 139 chemicals), the DPRA/ARE assay (N = 69), the SH test/ARE assay (N = 73), the h-CLAT/DPRA/ARE assay (N = 69) and the h-CLAT/SH test/ARE assay (N = 73). The h-CLAT/DPRA, h-CLAT/DPRA/ARE assay and h-CLAT/SH test/ARE assay combinations showed a better predictive performance than the DPRA/ARE assay and the SH test/ARE assay. Our data indicates that the descriptors evaluated in this study were all useful for predicting human skin sensitization potential, although combinations containing h-CLAT (reflecting dendritic cell-activating ability) were most effective for ANN-based prediction.

The Myeloid U937 Skin Sensitization Test (U-SENS) addresses the activation of dendritic cell event in the adverse outcome pathway for skin sensitization

The U-SENS™ assay, formerly known as MUSST (Myeloid U937 Skin Sensitization Test), is an in vitro method to assess skin sensitization. Dendritic cell activation following exposure to sensitizers was modelled in the U937 human myeloid cell line by measuring the induction of the expression of CD86 by flow cytometry. The predictive performance of U-SENS™ was assessed via a comprehensive comparison analysis with the available human and LLNA data of 175 substances. U-SENS™ showed 79% specificity, 90% sensitivity and 88% accuracy. A four laboratory ring study demonstrated the transferability, reliability and reproducibility of U-SENS™, with a reproducibility of 95% within laboratories and 79% between-laboratories, showing that the U-SENS™ assay is a promising tool in a skin sensitization risk assessment testing strategy.

Proposing a scientific confidence framework to help support the application of adverse outcome pathways for regulatory purposes

An adverse outcome pathway (AOP) describes the causal linkage between initial molecular events and an adverse outcome at individual or population levels. Whilst there has been considerable momentum in AOP development, far less attention has been paid to how AOPs might be practically applied for different regulatory purposes. This paper proposes a scientific confidence framework (SCF) for evaluating and applying a given AOP for different regulatory purposes ranging from prioritizing chemicals for further evaluation, to hazard prediction, and ultimately, risk assessment. The framework is illustrated using three different AOPs for several typical regulatory applications. The AOPs chosen are ones that have been recently developed and/or published, namely those for estrogenic effects, skin sensitisation, and rodent liver tumor promotion. The examples confirm how critical the data-richness of an AOP is for driving its practical application. In terms of performing risk assessment, human dosimetry methods are necessary to inform meaningful comparisons with human exposures; dosimetry is applied to effect levels based on non-testing approaches and in vitro data. Such a comparison is presented in the form of an exposure:activity ratio (EAR) to interpret biological activity in the context of exposure and to provide a basis for product stewardship and regulatory decision making.

Predicting chemically-induced skin reactions. Part II: QSAR models of skin permeability and the relationships between skin permeability and skin sensitization

Skin permeability is widely considered to be mechanistically implicated in chemically-induced skin sensitization. Although many chemicals have been identified as skin sensitizers, there have been very few reports analyzing the relationships between molecular structure and skin permeability of sensitizers and non-sensitizers. The goals of this study were to: (i) compile, curate, and integrate the largest publicly available dataset of chemicals studied for their skin permeability; (ii) develop and rigorously validate QSAR models to predict skin permeability; and (iii) explore the complex relationships between skin sensitization and skin permeability. Based on the largest publicly available dataset compiled in this study, we found no overall correlation between skin permeability and skin sensitization. In addition, cross-species correlation coefficient between human and rodent permeability data was found to be as low as R²=0.44. Human skin permeability models based on the random forest method have been developed and validated using OECD-compliant QSAR modeling workflow. Their external accuracy was high (Q²_ext = 0.73 for 63% of external compounds inside the applicability domain). The extended analysis using both experimentally-measured and QSAR-imputed data still confirmed the absence of any overall concordance between skin permeability and skin sensitization. This observation suggests that chemical modifications that affect skin permeability should not be presumed a priori to modulate the sensitization potential of chemicals. The models reported herein as well as those developed in the companion paper on skin sensitization suggest that it may be possible to rationally design compounds with the desired high skin permeability but low sensitization potential.

Assessment of the skin sensitising potency of the lower alkyl methacrylate esters

There is continued interest in, and imperatives for, the classification of contact allergens according to their relative skin sensitising potency. However, achieving that end can prove problematic, not least when there is an apparent lack of concordance between experimental assessments of potency and the prevalence allergic contact dermatitis as judged by clinical experience. For the purpose of exploring this issue, and illustrating the important considerations that are required to reach sound judgements about potency categorisation, the lower alkyl methacrylate esters (LAM) have been employed here as a case study. Although the sensitising potential of methyl methacrylate (MMA) has been reviewed previously, there is available new information that is relevant for assessment of skin sensitising potency. Moreover, for the purposes of this article, analyses have been extended to include also other LAM for which relevant data are available: ethyl methacrylate (EMA), n-butyl methacrylate (nBMA), isobutyl methacrylate (iBMA), and 2-ethylhexyl methacrylate (EHMA). In addressing the skin sensitising activity of these chemicals and in drawing conclusions regarding relative potency, a number of sources of information has been considered, including estimates of potency derived from local lymph node assay (LLNA) data, the results of guinea pig assays, and data derived from in silico methods and from recently developed in vitro approaches. Moreover, clinical experience of skin sensitisation of humans by LAM has also been evaluated. The conclusion drawn is that MMA and other LAM are contact allergens, but that none of these chemicals has any more than weak skin sensitising potency. We have also explored here the possible bases for this modest sensitising activity. Finally, the nature of exposure to LAM has been reviewed briefly and on the basis of that information, together with an understanding of skin sensitising potency, a risk assessment has been prepared.

Towards AOP application--implementation of an integrated approach to testing and assessment (IATA) into a pipeline tool for skin sensitization

Since the OECD published the Adverse Outcome Pathway (AOP) for skin sensitization, many efforts have focused on how to integrate and interpret nonstandard information generated for key events in a manner that can be practically useful for decision making. These types of frameworks are known as Integrated Approaches to Testing and Assessment (IATA). Here we have outlined an IATA for skin sensitization which focuses on existing information including non testing approaches such as QSAR and read-across. The IATA was implemented into a pipeline tool using OASIS technology to provide a means of systematically collating and compiling relevant information which could be used in an assessment of skin sensitization potential. A test set of 100 substances with available skin sensitization information was profiled using the pipeline IATA. In silico and in chemico profiling information alone was able to correctly predict skin sensitization potential, with a preliminary accuracy of 73.85%. Information from other relevant endpoints (e.g., Ames mutagenicity) was found to improve the accuracy (to 87.6%) when coupled with a reaction chemistry mechanistic understanding. This pipeline platform could be useful in the assessment of skin sensitization potential and marks a step change in how non testing approaches can be practically applied.

Evaluating the performance of integrated approaches for hazard identification of skin sensitizing chemicals

The currently available animal-free methods for the detection of skin sensitizing potential of chemicals seem promising. However, no single method is able to comprehensively represent the complexity of the processes involved in skin sensitization. To ensure a mechanistic basis and cover the complexity, multiple methods should be integrated into a testing strategy, in accordance with the adverse outcome pathway that describes all key events in skin sensitization. Although current majority voting testing strategies have proven effective, the performance of individual methods is not taken into account. To that end, we designed a tiered strategy based on complementary characteristics of the included methods, and compared it to a majority voting approach. This tiered testing strategy was able to correctly identify all 41 chemicals tested. In terms of total number of experiments required, the tiered testing strategy requires less experiments compared to the majority voting approach. On the other hand, this tiered strategy is more complex due the number of different alternative methods required, and predicted costs are similar for both strategies. Both the tiered and majority voting strategies provide a mechanistic basis for skin sensitization testing, but the strategy most suitable for regulatory decision-making remains to be determined.

Development of a multiparametric in vitro model of skin sensitization

Most animal experiments on cosmetics safety are prohibited and since March 2013, this obligation includes sensitization tests. However, until now there has been no validated alternative in vitro method. In this work, 400 compounds used in the cosmetic industry were selected to cover the greatest diversity of structures, biological activities and sensitizing potential. These molecules were submitted to a series of tests aimed at reproducing essential steps in sensitization and to distinguish between sensitization and irritations, i.e., transcutaneous permeation (factor A), haptenation (factor B), sensitization cytokines production (factor C) and acute toxicity (factor D). The transcutaneous diffusion was measured on human skin explants using Franz cells. Haptenation was tested in solution on human serum albumin. Sensitization cytokine production was investigated by measurement of interleukin-18 release by keratinocytes. Acute toxicity was determined using an 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide(75) cell viability test. As only sufficiently stable, soluble and detectable compounds are usable, 33, 72, 68 and 68 molecules were finally tested on factors A, B, C and D, respectively, and 32 were completely screened by the four factors. The individual correlation of the four factors with the reference in vivo tests was limited but the combination of these factors led to a correlation between in vivo and in vitro assays of 81.2% and the safety of the test (risk of false negative) reached 96.8%. The techniques employed are simple and inexpensive and this model of four tests appears as a promising technique to evaluate in vitro the skin sensitization potential of unknown molecules.

Skin sensitization risk assessment model using artificial neural network analysis of data from multiple in vitro assays

The sensitizing potential of chemicals is usually identified and characterized using in vivo methods such as the murine local lymph node assay (LLNA). Due to regulatory constraints and ethical concerns, alternatives to animal testing are needed to predict skin sensitization potential of chemicals. For this purpose, combined evaluation using multiple in vitro and in silico parameters that reflect different aspects of the sensitization process seems promising. We previously reported that LLNA thresholds could be well predicted by using an artificial neural network (ANN) model, designated iSENS ver.1 (integrating in vitro sensitization tests version 1), to analyze data obtained from two in vitro tests: the human Cell Line Activation Test (h-CLAT) and the SH test. Here, we present a more advanced ANN model, iSENS ver.2, which additionally utilizes the results of antioxidant response element (ARE) assay and the octanol-water partition coefficient (LogP, reflecting lipid solubility and skin absorption). We found a good correlation between predicted LLNA thresholds calculated by iSENS ver.2 and reported values. The predictive performance of iSENS ver.2 was superior to that of iSENS ver.1. We conclude that ANN analysis of data from multiple in vitro assays is a useful approach for risk assessment of chemicals for skin sensitization.

Integrated testing and assessment approaches for skin sensitization: a commentary

A Bayesian integrated testing strategy (ITS) approach, aiming to assess skin sensitization potency, has been presented, in which data from various types of in vitro assays are integrated and assessed in combination for their ability to predict in vivo skin sensitization data. Here we discuss this approach and compare it to our quantitative mechanistic modeling (QMM) approach based on physical organic chemistry. The main findings of the Bayesian study are consistent with our chemistry-based approach and our previously published assessment of the key determinants of sensitization potency, in particular the relatively high predictive value found for chemical reactivity data and the relatively low predictive value for bioavailability parameters. As it stands at present the Bayesian approach does not utilize the full range of predictive capability that is already available, and aims only to assign potency categories rather than numerical potency values per se. In contrast, for many chemicals the QMM approach can already provide numerical potency predictions. However, the Bayesian approach may have potential for those chemicals where a chemistry modeling approach cannot provide a complete answer (e.g. pro-electrophiles whose in cutaneo activation cannot currently be modeled confidently). Nonetheless, our main message is of the importance of leveraging chemistry insights and read-across approaches to the fullest extent possible.

A dataset on 145 chemicals tested in alternative assays for skin sensitization undergoing prevalidation

Skin sensitization is a key endpoint for cosmetic ingredients, with a forthcoming ban for animal testing in Europe. Four alternative tests have so far been submitted to ECVAM prevalidation: (i) MUSST and (ii) h-Clat assess surface markers on dendritic cell lines, (iii) the direct peptide reactivity assay (DPRA) measures reactivity with model peptides and (iv) the KeratinoSens(TM) assay which is based on detection of Nrf2-induced luciferase. It is anticipated that only an integrated testing strategy (ITS) based on a battery of tests might give a full replacement providing also a sensitization potency assessment, but this concept should be tested with a data-driven analysis. Here we report a database on 145 chemicals reporting the quantitative endpoints measured in a U937- test, the DPRA and KeratinoSens(TM) . It can serve to develop data-driven ITS approaches as we show in a parallel paper and provides a view as to the current ability to predict with in vitro tests as we are entering 2013. It may also serve as reference database when benchmarking new molecules with in vitro based read-across and find use as a reference database when evaluating new tests. The tests and combinations thereof were evaluated for predictivity, and overall a similar predictivity was found as before on three-fold smaller datasets. Analysis of the dose-response parameters of the individual tests indicates a correlation to sensitization potency. Detailed analysis of chemicals false-negative and false-positive in two tests helped to define limitations in the tests but also in the database derived from animal studies.