You are here:
Evaluating an In Vitro Distribution Model
Citation:
Scherer, M., K. Paul-Friedman, AND J. Wambaugh. Evaluating an In Vitro Distribution Model. Society of Toxicology (SOT) 63rd Annual Meeting and ToxExpo, Salt Lake City, UT, March 10 - 14, 2024. https://doi.org/10.23645/epacomptox.25398871
Impact/Purpose:
Presentation to the Society of Toxicology (SOT) 63rd Annual Meeting and ToxExpo March 2024
Description:
Background and Purpose: To establish toxicity reference values, risk assessments require reliable methods with well-characterized uncertainty. Next generation risk assessment will require in vitro to in vivo extrapolation (IVIVE) to translate observed cellular responses to whole organisms. IVIVE informs quantitative dose-response using mathematical modeling and in vitro bioactivity assay data. In vitro disposition is an important part of IVIVE and refers to the way that a given chemical partitions within the in vitro system via binding to the plate wall, media, proteins, cells, and volatilization to air. The distribution of a chemical dictates the difference between the nominal and bioavailable chemical concentration that causes any observed effects. Chemical distribution varies across chemicals as a function of both inherent chemical properties and in vitro test conditions. Given that in vitro bioactivity screening has been performed across large chemical libraries (for example, ToxCast and Tox21), there is a critical need to accurately predict in vitro distribution for many chemicals. Currently, most IVIVE models only use the nominal values. Methods: The available literature was evaluated to find studies that reported experimentally derived intracellular concentrations from in vitro tests. Information characterizing the experimental conditions was then input to the Armitage et al. (2014) in vitro disposition model as implemented within R package “httk” and the Kramer et al. (2010) in vitro disposition model. The results from these models were compared using root mean squared log error (RMSLE) to assess how well the model predicted the concentrations. Results: Generally, the Armitage model predicts the intracellular concentrations to be up to 4-fold larger than their experimental value. The Kramer model is more accurate, predicting the intracellular concentrations with a smaller spread: only up to 2-fold different than their experimental value. Both models are able to capture trends in intracellular concentrations for chemicals with multiple observations. Among the four chemicals with multiple observations, the Kramer model most accurately predicted positive chemicals, overpredicted negative chemicals, and underpredicted neutral chemicals. The Armitage model most accurately predicted neutral chemicals, slightly overpredicted positive chemicals, and largely overpredicted the negative chemicals. Conclusions: Models that are relying solely on the Armitage or Kramer model to predict intracellular concentrations could be underpredicting the bioactive dose from in vitro to in vivo extrapolation. This would have ramifications for both risk assessment and decision making. Ultimately, a more nuanced method is needed, using either the Armitage or Kramer models based on the goals of the project and corresponding to the properties of the chemicals used (i.e., positively or negatively charged at physiological pH). The literature review only returned 55 data points (i.e., pairs of measured and experimental intracellular concentrations). This makes it difficult to draw definitive conclusions. In the future, a specific assay to generate data for in vitro distribution models would aid in the evaluation of mathematical models for the prediction of intracellular concentrations.
URLs/Downloads:
DOI: Evaluating an In Vitro Distribution Model
SOT_POSTER_FINAL_03012024.PDF (PDF, NA pp, 542.694 KB, about PDF)