Computational Tools Applications

ICCVAM and its member agencies are exploring how computational approaches can be applied to reduce animal use for toxicity testing. These approaches have potential application for acute oral toxicity and eye irritation testing, and for predicting whether chemicals could cause cardiotoxicity, developmental toxicity, or neurotoxicity.

CATMoS and Additivity Approaches to Predict Toxicity of Mixtures

The majority of existing toxicity data and tools to assess toxicity are for single compounds. However, humans are exposed to complex chemical mixtures in a variety of scenarios. Of particular interest to the U.S. Air Force are exposures to chemical mixtures experienced by airmen. Scientists at the U.S. Air Force Research Laboratory have developed a computational approach to help assess the acute oral toxicity of mixtures by using an additivity formula that estimates toxicity of a mixture based on the acute toxicity estimate of individual ingredients. Acute toxicity estimates for single chemicals were collected from existing databases or calculated using CATMoS (Chushak et al. 2021). The proposed approach was implemented in a standalone Python-based app that allows users to rapidly assess the acute toxicity for mixtures related to U.S. Air Force needs and to identify chemicals that can impact physiological performance of airmen.

Analysis of the Performance of the GHS Mixtures Equation to Predict Acute Oral Toxicity of Formulations

The majority of pesticide registration applications require product-specific acute toxicity data. Thus, an alternative to in vivo testing for this purpose could greatly reduce animal testing. However, predicting toxicity using NAMs is challenging for mixtures such as pesticides. The GHS Mixtures Equation estimates the acute toxicity of mixtures using the toxicities of mixture components. EPA and NIEHS scientists conducted a study (Hamm et al. 2021) to evaluate the concordance of hazard classifications predicted using the GHS Mixtures Equation with classifications based on in vivo test results. The results of the analysis suggested that the GHS Mixtures Equation can help predict the acute oral toxicity of mixtures, particularly those with lower toxicity.

Use of IVIVE and Data from a Stem Cell-based Assay to Predict Developmental Toxicity Potential

To support implementation of NAMs for regulatory decision-making on developmental toxicity, FDA and NIEHS scientists and collaborators evaluated the performance of the devTOX quickPredict assay for predicting lowest effect levels in rat developmental toxicity studies. Studies conducted during 2020 and 2021 focused on developmental toxicity potential of valproic acid and analogues (Chang et al. 2022) and 186 chemicals from the Tox21 program. Developmental toxicity potential (dTP) concentrations from the devTOX quickPredict assay were used as inputs to IVIVE analyses to estimate equivalent administered doses that would result in the maximum plasma concentrations equivalent to dTP concentrations. Results suggested that the devTOX quickPredict assay can quantitatively predict developmental toxicity potential at concentrations relevant to human exposure and in some cases may provide a more conservative hazard estimate than animal studies for use in risk assessment.

In Silico Screening Approaches for Assessing Cardiovascular Safety

Mounting evidence supports the contribution of environmental chemicals to cardiovascular disease burden. NIEHS scientists evaluated chemicals in the Tox21 chemical library for cardiotoxicity potential by focusing on HTS assays whose targets are associated with adverse events related to cardiovascular failure modes (Krishna et al. 2021). The objective of the evaluation was to develop new hypotheses around environmental chemicals of potential interest for adverse cardiovascular outcomes using Tox21 and ToxCast HTS data. Bioactivity signatures relevant to cardiotoxicity were identified for 40 targets measured in 314 assays and used to prioritize 1,138 Tox21 chemicals. The approach identified drugs with known cardiotoxic effects in a variety of use classes including estrogenic modulators, anti-arrhythmic drugs, and antipsychotic drugs like chlorpromazine. Several classes of environmental chemicals such as organotins, bisphenol-like chemicals, pesticides, and quaternary ammonium compounds demonstrated strong bioactivity against cardiovascular targets. Screening outcomes were added to existing data from literature studies using cultured heart cells, animals, or human epidemiological approaches to prioritize these chemicals for further testing.

A number of chemicals identified in the initial Tox21 screen were found to inhibit potassium channels important to cardiac rhythm regulation (Krishna et al. 2022). A set of molecular descriptors was applied to characterize these chemicals. Machine learning approaches were then applied to build robust statistical models that can predict the probability of any new chemical to cause cardiotoxicity via this mechanism.

Application of CATMoS to Predict Acute Mammalian Toxicity for Pesticide Hazard and Risk Assessment

CATMoS is a set of predictive in silico models of acute oral systemic toxicity potential. The consensus model predictions are fully reproducible and demonstrate equivalent performance to in vivo data. This offers an opportunity for a potential replacement to animal testing for applications such as regulatory evaluations of pesticide toxicity. EPA and NIEHS scientists are evaluating CATMoS predictions for 178 chemicals in comparison to rat LD50 tests from publicly available ecological risk assessments registered from 1998 to 2020. Findings from this study will help in understanding the applicability of CATMoS estimates as a potential replacement of the rat acute single oral dose study for establishing the effects endpoint in ecological risk assessments.