Transferability and reproducibility of the EpiSkin™ Micronucleus Assay.

A novel in vitro 3D micronucleus assay was developed in China using the EpiSkin™ 3D human skin model. This EpiSkin™ Micronucleus Assay showed good predictivity and reproducibility during internal validation and is expected to contribute to in vitro genotoxicity testing as a follow-up for positive results from 2D micronucleus assay. Having developed the assay in one laboratory, further work focused on the transferability and inter-laboratory reproducibility in two additional Chinese authority laboratories (Guangdong Provincial Center for Disease Control and Prevention and Zhejiang Institute for Food and Drug Control). Formal training was provided for both laboratories, which resulted in good transferability based on the results of two positive compounds, such as mitomycin C and vinblastine. Independent experiments were then performed, and inter-laboratory reproducibility was checked using 2-acetylaminofluorene, 5-fluorouracil, 2,4-dichlorophenol, and d-limonene. The dose-responses of the positive control chemical, mitomycin C, were similar to those of the developing laboratory, and all test chemicals were correctly classified by all laboratories. Overall, there was a good transferability as well as intra- and inter-laboratory reproducibility of the EpiSkin™ Micronucleus Assay. This study further confirmed the assay's robustness and provided confidence to enter following validation stages for scientific acceptance.

Read-across and new approach methodologies applied in a 10-step framework for cosmetics safety assessment - A case study with parabens.

Parabens are esters of para-hydroxybenzoic acid that have been used as preservatives in many types of products for decades including agrochemicals, pharmaceuticals, food and cosmetics. This illustrative case study with propylparaben (PP) demonstrates a 10-step read-across (RAX) framework in practice. It aims at establishing a proof-of-concept for the value added by new approach methodologies (NAMs) in read-across (RAX) for use in a next-generation risk assessment (NGRA) in order to assess consumer safety after exposure to PP-containing cosmetics. In addition to structural and physico-chemical properties, in silico information, toxicogenomics, in vitro toxicodynamic, toxicokinetic data from PBK models, and bioactivity data are used to provide evidence of the chemical and biological similarity of PP and analogues and to establish potency trends for observed effects in vitro. The chemical category under consideration is short (C1-C4) linear chain n-alkyl parabens: methylparaben, ethylparaben, propylparaben and butylparaben. The goal of this case study is to illustrate how a practical framework for RAX can be used to fill a hypothetical data gap for reproductive toxicity of the target chemical PP.

Use of in vitro metabolism and biokinetics assays to refine predicted in vivo and in vitro internal exposure to the cosmetic ingredient, phenoxyethanol, for use in risk assessment.

A novel approach was developed to help characterize the biokinetics of the cosmetic ingredient, phenoxyethanol, to help assess the safety of the parent and its major stable metabolite. In the first step of this non-animal tiered approach, primary human hepatocytes were used to confirm or refute in silico predicted metabolites, and elucidate the intrinsic clearance of phenoxyethanol. A key result was the identification of the major metabolite, phenoxyacetic acid (PAA), the exposure to which in the kidney was subsequently predicted to far exceed that of phenoxyethanol in blood or other tissues. Therefore, a novel aspect of this approach was to measure in the subsequent step the formation of PAA in the cells dosed with phenoxyethanol that were used to provide points of departure (PoDs) and express the intracellular exposure as the Cmax and AUC24. This enabled the calculation of the intracellular concentrations of parent and metabolite at the PoD in the cells used to derive this value. These concentrations can be compared with in vivo tissue levels to conclude on the safety margin. The lessons from this case study will help to inform the design of other non-animal safety assessments.

A 10-step framework for use of read-across (RAX) in next generation risk assessment (NGRA) for cosmetics safety assessment.

This paper presents a 10-step read-across (RAX) framework for use in cases where a threshold of toxicological concern (TTC) approach to cosmetics safety assessment is not possible. RAX builds on established approaches that have existed for more than two decades using chemical properties and in silico toxicology predictions, by further substantiating hypotheses on toxicological similarity of substances, and integrating new approach methodologies (NAM) in the biological and kinetic domains. NAM include new types of data on biological observations from, for example, in vitro assays, toxicogenomics, metabolomics, receptor binding screens and uses physiologically-based kinetic (PBK) modelling to inform about systemic exposure. NAM data can help to substantiate a mode/mechanism of action (MoA), and if similar chemicals can be shown to work by a similar MoA, a next generation risk assessment (NGRA) may be performed with acceptable confidence for a data-poor target substance with no or inadequate safety data, based on RAX approaches using data-rich analogue(s), and taking account of potency or kinetic/dynamic differences.

Utility of a next-generation framework for assessment of genomic damage: A case study using the pharmaceutical drug candidate etoposide.

We present a hypothetical case study to examine the use of a next-generation framework developed by the Genetic Toxicology Technical Committee of the Health and Environmental Sciences Institute for assessing the potential risk of genetic damage from a pharmaceutical perspective. We used etoposide, a genotoxic carcinogen, as a representative pharmaceutical for the purposes of this case study. Using the framework as guidance, we formulated a hypothetical scenario for the use of etoposide to illustrate the application of the framework to pharmaceuticals. We collected available data on etoposide considered relevant for assessment of genetic toxicity risk. From the data collected, we conducted a quantitative analysis to estimate margins of exposure (MOEs) to characterize the risk of genetic damage that could be used for decision-making regarding the predefined hypothetical use. We found the framework useful for guiding the selection of appropriate tests and selecting relevant endpoints that reflected the potential for genetic damage in patients. The risk characterization, presented as MOEs, allows decision makers to discern how much benefit is critical to balance any adverse effect(s) that may be induced by the pharmaceutical. Interestingly, pharmaceutical development already incorporates several aspects of the framework per regulations and health authority expectations. Moreover, we observed that quality dose response data can be obtained with carefully planned but routinely conducted genetic toxicity testing. This case study demonstrates the utility of the next-generation framework to quantitatively model human risk based on genetic damage, as applicable to pharmaceuticals.

New framework for a non-animal approach adequately assures the safety of cosmetic ingredients - A case study on caffeine.

This case study on the model substance caffeine demonstrates the viability of a 10-step read-across (RAX) framework in practice. New approach methodologies (NAM), including RAX and physiologically-based kinetic (PBK) modelling were used to assess the consumer safety of caffeine. Appropriate animal systemic toxicity data were used from the most relevant RAX analogue while assuming that no suitable animal toxicity data were available for caffeine. Based on structural similarities, three primary metabolites of the target chemical caffeine (theophylline, theobromine and paraxanthine) were selected as its most relevant analogues, to estimate a point of departure in order to support a next generation risk assessment (NGRA). On the basis of the pivotal mode of action (MOA) of caffeine and other methylxanthines, theophylline appeared to be the most potent and suitable analogue. A worst-case aggregate exposure assessment determined consumer exposure to caffeine from different sources, such as cosmetics and food/drinks. Using a PBK model to estimate human blood concentrations following exposure to caffeine, an acceptable Margin of Internal Exposure (MOIE) of 27-fold was derived on the basis of a RAX using theophylline animal data, which suggests that the NGRA approach for caffeine is sufficiently conservative to protect human health.

Validation of the 3D reconstructed human skin micronucleus (RSMN) assay: an animal-free alternative for following-up positive results from standard in vitro genotoxicity assays.

In vitro test batteries have become the standard approach to determine the genotoxic potential of substances of interest across industry sectors. While useful for hazard identification, standard in vitro genotoxicity assays in 2D cell cultures have limited capability to predict in vivo outcomes and may trigger unnecessary follow-up animal studies or the loss of promising substances where animal tests are prohibited or not desired. To address this problem, a team of regulatory, academia and industry scientists was established to develop and validate 3D in vitro human skin-based genotoxicity assays for use in testing substances with primarily topical exposure. Validation of the reconstructed human skin micronucleus (RSMN) assay in MatTek Epi-200™ skin models involved testing 43 coded chemicals selected by independent experts, in four US/European laboratories. The results were analysed by an independent statistician according to predefined criteria. The RSMN assay showed a reproducibly low background micronucleus frequency and exhibited sufficient capacity to metabolise pro-mutagens. The overall RSMN accuracy when compared to in vivo genotoxicity outcomes was 80%, with a sensitivity of 75% and a specificity of 84%, and the between- and within-laboratory reproducibility was 77 and 84%, respectively. A protocol involving a 72-h exposure showed increased sensitivity in detecting true positive chemicals compared to a 48-h exposure. An analysis of a test strategy using the RSMN assay as a follow-up test for substances positive in standard in vitro clastogenicity/aneugenicity assays and a reconstructed skin Comet assay for substances with positive results in standard gene mutation assays results in a sensitivity of 89%. Based on these results, the RSMN assay is considered sufficiently validated to establish it as a 'tier 2' assay for dermally exposed compounds and was recently accepted into the OECD's test guideline development program.

A new 3D model for genotoxicity assessment: EpiSkin™ Micronucleus Assay.

The European Regulation on Cosmetics (no. 1223/2009) has prohibited the use of animals in safety testing since March 2009 for ingredients used in cosmetics. Irreversible events at the chromosome level (clastogenesis and aneugenesis) are commonly evaluated by scoring either micronuclei or chromosome aberrations using cell-based genotoxicity assays. Like most in vitro genotoxicity assays, the 2D in vitro micronucleus assay exhibits a poor specificity and does not mimic the dermal route. To address these limitations, the current project aims to develop and validate a 3D micronucleus assay using the EpiSkin™ model. This project is scientifically supported by the Cosmetics Europe Genotoxicity Task Force. In a first step, two key criteria for the development of micronucleus assay, namely, the sufficient yield of cells from the EpiSkin™ model and an acceptable proliferation rate of the basal layer, were assessed and demonstrated. Subsequently, six chemicals (vinblastine, n-ethylnitrosourea, ß-butyrolactone, 2-acetylaminofluorene, 2,4-dichlorophenoland d-limonene) were evaluated in the EpiSkin™ Micronucleus Assay. At least two independent experiments using 48- and 72-h incubations were performed for each chemical. Results showed good inter-experimental reproducibility, as well as the correct identification of all six tested chemicals. The metabolism of 2-acetylaminofluorene on the EpiSkin™ model was also investigated and confirmed by the formation of an intermediate metabolite (2-aminofluorene). These preliminary results from the EpiSkin™ Micronucleus Assay indicate that it is a promising in vitro assay for assessing genotoxicity. The availability and suitability of this test method contribute significantly to the development of non-animal testing methods in China and its impact on the worldwide field.

Use of in vitro 3D tissue models in genotoxicity testing: Strategic fit, validation status and way forward. Report of the working group from the 7th International Workshop on Genotoxicity Testing (IWGT).

Use of three-dimensional (3D) tissue equivalents in toxicology has been increasing over the last decade as novel preclinical test systems and as alternatives to animal testing. In the area of genetic toxicology, progress has been made with establishing robust protocols for skin, airway (lung) and liver tissue equivalents. In light of these advancements, a "Use of 3D Tissues in Genotoxicity Testing" working group (WG) met at the 7th IWGT meeting in Tokyo in November 2017 to discuss progress with these models and how they may fit into a genotoxicity testing strategy. The workshop demonstrated that skin models have reached an advanced state of validation following over 10 years of development, while liver and airway model-based genotoxicity assays show promise but are at an early stage of development. Further effort in liver and airway model-based assays is needed to address the lack of coverage of the three main endpoints of genotoxicity (mutagenicity, clastogenicity and aneugenicity), and information on metabolic competence. The IWGT WG believes that the 3D skin comet and micronucleus assays are now sufficiently validated to undergo an independent peer review of the validation study, followed by development of individual OECD Test Guidelines.

The way forward for assessing the human health safety of cosmetics in the EU - Workshop proceedings.

Although the need for non-animal alternatives has been well recognised for the human health hazard assessment of chemicals in general, it has become especially pressing for cosmetic ingredients due to the full implementation of testing and marketing bans on animal testing under the European Cosmetics Regulation. This means that for the safety assessment of cosmetics, the necessary safety data for both the ingredients and the finished product can be drawn from validated (or scientifically-valid), so-called "Replacement methods". In view of the challenges for safety assessment without recourse to animal test data, the Methodology Working Group of the Scientific Committee on Consumer Safety organised a workshop in February 2019 to discuss the key issues in regard to the use of animal-free alternative methods for the safety evaluation of cosmetic ingredients. This perspective article summarises the outcomes of this workshop and reflects on the state-of-the-art and possible way forward for the safety assessment of cosmetic ingredients for which no experimental animal data exist. The use and optimisation of "New Approach Methodology" that could be useful tools in the context of the "Next Generation Risk Assessment" and the strategic framework for safety assessment of cosmetics were discussed in depth.

Application of the adverse outcome pathway framework to genotoxic modes of action.

In May 2017, the Health and Environmental Sciences Institute's Genetic Toxicology Technical Committee hosted a workshop to discuss whether mode of action (MOA) investigation is enhanced through the application of the adverse outcome pathway (AOP) framework. As AOPs are a relatively new approach in genetic toxicology, this report describes how AOPs could be harnessed to advance MOA analysis of genotoxicity pathways using five example case studies. Each of these genetic toxicology AOPs proposed for further development includes the relevant molecular initiating events, key events, and adverse outcomes (AOs), identification and/or further development of the appropriate assays to link an agent to these events, and discussion regarding the biological plausibility of the proposed AOP. A key difference between these proposed genetic toxicology AOPs versus traditional AOPs is that the AO is a genetic toxicology endpoint of potential significance in risk characterization, in contrast to an adverse state of an organism or a population. The first two detailed case studies describe provisional AOPs for aurora kinase inhibition and tubulin binding, leading to the common AO of aneuploidy. The remaining three case studies highlight provisional AOPs that lead to chromosome breakage or mutation via indirect DNA interaction (inhibition of topoisomerase II, production of cellular reactive oxygen species, and inhibition of DNA synthesis). These case studies serve as starting points for genotoxicity AOPs that could ultimately be published and utilized by the broader toxicology community and illustrate the practical considerations and evidence required to formalize such AOPs so that they may be applied to genetic toxicity evaluation schemes. Environ. Mol. Mutagen. 61:114-134, 2020. © 2019 Wiley Periodicals, Inc.

Towards grouping concepts based on new approach methodologies in chemical hazard assessment: the read-across approach of the EU-ToxRisk project.

Read-across is one of the most frequently used alternative tools for hazard assessment, in particular for complex endpoints such as repeated dose or developmental and reproductive toxicity. Read-across extrapolates the outcome of a specific toxicological in vivo endpoint from tested (source) compounds to "similar" (target) compound(s). If appropriately applied, a read-across approach can be used instead of de novo animal testing. The read-across approach starts with structural/physicochemical similarity between target and source compounds, assuming that similar structural characteristics lead to similar human hazards. In addition, similarity also has to be shown for the toxicokinetic and toxicodynamic properties of the grouped compounds. To date, many read-across cases fail to demonstrate toxicokinetic and toxicodynamic similarities. New concepts, in vitro and in silico tools are needed to better characterise these properties, collectively called new approach methodologies (NAMs). This white paper outlines a general read-across assessment concept using NAMs to support hazard characterization of the grouped compounds by generating data on their dynamic and kinetic properties. Based on the overarching read-across hypothesis, the read-across workflow suggests targeted or untargeted NAM testing also outlining how mechanistic knowledge such as adverse outcome pathways (AOPs) can be utilized. Toxicokinetic models (biokinetic and PBPK), enriched by in vitro parameters such as plasma protein binding and hepatocellular clearance, are proposed to show (dis)similarity of target and source compound toxicokinetics. Furthermore, in vitro to in vivo extrapolation is proposed to predict a human equivalent dose, as potential point of departure for risk assessment. Finally, the generated NAM data are anchored to the existing in vivo data of source compounds to predict the hazard of the target compound in a qualitative and/or quantitative manner. To build this EU-ToxRisk read-across concept, case studies have been conducted and discussed with the regulatory community. These case studies are briefly outlined.

Profiling 58 compounds including cosmetic-relevant chemicals using ToxRefDB and ToxCast.

Safety assessment for cosmetic-relevant chemicals (CRCs) in the European Union has been reshaped by restrictions on animal testing, and new approach methodologies (NAMs) for predicting toxicity are critical to ensure new cosmetic product safety. To demonstrate NAMs for safety assessment, we surveyed in vitro bioactivity and in vivo systemic toxicity data in the US Environmental Protection Agency's (EPA's) Toxicity Forecaster (ToxCast) and Toxicity Reference databases (ToxRefDB), respectively, for 58 chemicals identified as CRCs, including cosmetic ingredients as well as trace contaminants. CRCs were diverse in use types as suggested by broad chemical use categories. In terms of both target organ effects and study type, the median of the lowest effect level (LEL) doses in ToxRefDB for CRCs tended to be slightly higher than the median for the remaining 928 chemicals with study data in ToxRefDB, though the ranges of LELs were similar. For 17 of the 58 CRCs, high-throughput toxicokinetic data were used to calculate administered equivalent doses (AEDs) in mg/kg/day units for the in vitro bioactivity observed, and these AEDs served as conservative estimators of the systemic LELs observed in vivo. This work suggests that NAMs for bioactivity may inform a conservative point-of-departure estimate for diverse CRCs.

Challenges and Opportunities in the Design of Liver-on-Chip Microdevices.

The liver is the central hub of xenobiotic metabolism and consequently the organ most prone to cosmetic- and drug-induced toxicity. Failure to detect liver toxicity or to assess compound clearance during product development is a major cause of postmarketing product withdrawal, with disastrous clinical and financial consequences. While small animals are still the preferred model in drug development, the recent ban on animal use in the European Union created a pressing need to develop precise and efficient tools to detect human liver toxicity during cosmetic development. This article includes a brief review of liver development, organization, and function and focuses on the state of the art of long-term cell culture, including hepatocyte cell sources, heterotypic cell-cell interactions, oxygen demands, and culture medium formulation. Finally, the article reviews emerging liver-on-chip devices and discusses the advantages and pitfalls of individual designs. The goal of this review is to provide a framework to design liver-on-chip devices and criteria with which to evaluate this emerging technology.

A mode-of-action ontology model for safety evaluation of chemicals: Outcome of a series of workshops on repeated dose toxicity.

Repeated dose toxicity evaluation aims at assessing the occurrence of adverse effects following chronic or repeated exposure to chemicals. Non-animal approaches have gained importance in the last decades because of ethical considerations as well as due to scientific reasons calling for more human-based strategies. A critical aspect of this challenge is linked to the capacity to cover a comprehensive set of interdependent mechanisms of action, link them to adverse effects and interpret their probability to be triggered in the light of the exposure at the (sub)cellular level. Inherent to its structured nature, an ontology addressing repeated dose toxicity could be a scientific and transparent way to achieve this goal. Additionally, repeated dose toxicity evaluation through the use of a harmonized ontology should be performed in a reproducible and consistent manner, while mimicking as accurately as possible human physiology and adaptivity. In this paper, the outcome of a series of workshops organized by Cosmetics Europe on this topic is reported. As such, this manuscript shows how experts set critical elements and ways of establishing a mode-of-action ontology model as a support to risk assessors aiming to perform animal-free safety evaluation of chemicals based on repeated dose toxicity data.

Microphysiological flux balance platform unravels the dynamics of drug induced steatosis.

Drug development is currently hampered by the inability of animal experiments to accurately predict human response. While emerging organ on chip technology offers to reduce risk using microfluidic models of human tissues, the technology still mostly relies on end-point assays and biomarker measurements to assess tissue damage resulting in limited mechanistic information and difficulties to detect adverse effects occurring below the threshold of cellular damage. Here we present a sensor-integrated liver on chip array in which oxygen is monitored using two-frequency phase modulation of tissue-embedded microprobes, while glucose, lactate and temperature are measured in real time using microfluidic electrochemical sensors. Our microphysiological platform permits the calculation of dynamic changes in metabolic fluxes around central carbon metabolism, producing a unique metabolic fingerprint of the liver's response to stimuli. Using our platform, we studied the dynamics of human liver response to the epilepsy drug Valproate (Depakine™) and the antiretroviral medication Stavudine (Zerit™). Using E6/E7LOW hepatocytes, we show TC50 of 2.5 and 0.8 mM, respectively, coupled with a significant induction of steatosis in 2D and 3D cultures. Time to onset analysis showed slow progressive damage starting only 15-20 hours post-exposure. However, flux analysis showed a rapid disruption of metabolic homeostasis occurring below the threshold of cellular damage. While Valproate exposure led to a sustained 15% increase in lipogenesis followed by mitochondrial stress, Stavudine exposure showed only a transient increase in lipogenesis suggesting disruption of ß-oxidation. Our data demonstrates the importance of tracking metabolic stress as a predictor of clinical outcome.

Genotoxicity Assessment of Nanomaterials: Recommendations on Best Practices, Assays, and Methods.

Nanomaterials (NMs) present unique challenges in safety evaluation. An international working group, the Genetic Toxicology Technical Committee of the International Life Sciences Institute's Health and Environmental Sciences Institute, has addressed issues related to the genotoxicity assessment of NMs. A critical review of published data has been followed by recommendations on methods alterations and best practices for the standard genotoxicity assays: bacterial reverse mutation (Ames); in vitro mammalian assays for mutations, chromosomal aberrations, micronucleus induction, or DNA strand breaks (comet); and in vivo assays for genetic damage (micronucleus, comet and transgenic mutation assays). The analysis found a great diversity of tests and systems used for in vitro assays; many did not meet criteria for a valid test, and/or did not use validated cells and methods in the Organization for Economic Co-operation and Development Test Guidelines, and so these results could not be interpreted. In vivo assays were less common but better performed. It was not possible to develop conclusions on test system agreement, NM activity, or mechanism of action. However, the limited responses observed for most NMs were consistent with indirect genotoxic effects, rather than direct interaction of NMs with DNA. We propose a revised genotoxicity test battery for NMs that includes in vitro mammalian cell mutagenicity and clastogenicity assessments; in vivo assessments would be added only if warranted by information on specific organ exposure or sequestration of NMs. The bacterial assays are generally uninformative for NMs due to limited particle uptake and possible lack of mechanistic relevance, and are thus omitted in our recommended test battery for NM assessment. Recommendations include NM characterization in the test medium, verification of uptake into target cells, and limited assay-specific methods alterations to avoid interference with uptake or endpoint analysis. These recommendations are summarized in a Roadmap guideline for testing.

A strategy for systemic toxicity assessment based on non-animal approaches: The Cosmetics Europe Long Range Science Strategy programme.

When performing safety assessment of chemicals, the evaluation of their systemic toxicity based only on non-animal approaches is a challenging objective. The Safety Evaluation Ultimately Replacing Animal Test programme (SEURAT-1) addressed this question from 2011 to 2015 and showed that further research and development of adequate tools in toxicokinetic and toxicodynamic are required for performing non-animal safety assessments. It also showed how to implement tools like thresholds of toxicological concern (TTCs) and read-across in this context. This paper shows a tiered scientific workflow and how each tier addresses the four steps of the risk assessment paradigm. Cosmetics Europe established its Long Range Science Strategy (LRSS) programme, running from 2016 to 2020, based on the outcomes of SEURAT-1 to implement this workflow. Dedicated specific projects address each step of this workflow, which is introduced here. It tackles the question of evaluating the internal dose when systemic exposure happens. The applicability of the workflow will be shown through a series of case studies, which will be published separately. Even if the LRSS puts the emphasis on safety assessment of cosmetic relevant chemicals, it remains applicable to any type of chemical.

Predicting in vivo effect levels for repeat-dose systemic toxicity using chemical, biological, kinetic and study covariates.

In an effort to address a major challenge in chemical safety assessment, alternative approaches for characterizing systemic effect levels, a predictive model was developed. Systemic effect levels were curated from ToxRefDB, HESS-DB and COSMOS-DB from numerous study types totaling 4379 in vivo studies for 1247 chemicals. Observed systemic effects in mammalian models are a complex function of chemical dynamics, kinetics, and inter- and intra-individual variability. To address this complex problem, systemic effect levels were modeled at the study-level by leveraging study covariates (e.g., study type, strain, administration route) in addition to multiple descriptor sets, including chemical (ToxPrint, PaDEL, and Physchem), biological (ToxCast), and kinetic descriptors. Using random forest modeling with cross-validation and external validation procedures, study-level covariates alone accounted for approximately 15% of the variance reducing the root mean squared error (RMSE) from 0.96 log10 to 0.85 log10 mg/kg/day, providing a baseline performance metric (lower expectation of model performance). A consensus model developed using a combination of study-level covariates, chemical, biological, and kinetic descriptors explained a total of 43% of the variance with an RMSE of 0.69 log10 mg/kg/day. A benchmark model (upper expectation of model performance) was also developed with an RMSE of 0.5 log10 mg/kg/day by incorporating study-level covariates and the mean effect level per chemical. To achieve a representative chemical-level prediction, the minimum study-level predicted and observed effect level per chemical were compared reducing the RMSE from 1.0 to 0.73 log10 mg/kg/day, equivalent to 87% of predictions falling within an order-of-magnitude of the observed value. Although biological descriptors did not improve model performance, the final model was enriched for biological descriptors that indicated xenobiotic metabolism gene expression, oxidative stress, and cytotoxicity, demonstrating the importance of accounting for kinetics and non-specific bioactivity in predicting systemic effect levels. Herein, we generated an externally predictive model of systemic effect levels for use as a safety assessment tool and have generated forward predictions for over 30,000 chemicals.