Characterizing Accuracy of Model Predictions for Chemical Concentration in High Throughput Screening Assays

Area of Science:

• Toxicology
• Pharmacokinetics
• Computational Chemistry

Background:

• In vitro-in vivo extrapolation (IVIVE) aims to predict whole animal effects from in vitro data.
• Current IVIVE methods often rely on nominal concentrations, neglecting chemical distribution within assay systems.
• This can lead to inaccurate estimations of intracellular concentrations and, consequently, unreliable toxicity predictions.

Purpose of the Study:

• To evaluate the accuracy of mathematical disposition models in predicting intracellular chemical concentrations for IVIVE.
• To compare the performance of the Armitage et al. (2021) and Kramer et al. (2010) models against nominal concentrations.

Main Methods:

• Utilized 153 experimental intracellular concentration measurements for 43 chemicals from 12 studies and new data.
• Evaluated two in vitro disposition models: Armitage et al. (2021) and Kramer et al. (2010).
• Compared model predictions against experimentally determined intracellular concentrations and nominal concentrations.

Main Results:

• Both the Armitage model (RMSLE = 1.12) and Kramer model (RMSLE = 1.30) provided more accurate intracellular concentration predictions than nominal concentration (RMSLE = 1.45).
• The models demonstrated improved accuracy in estimating the free concentration of chemicals available to cause effects in vitro.

Conclusions:

• Mathematical modeling of in vitro chemical distribution significantly enhances the accuracy of IVIVE for toxicological assessments.
• These disposition models offer a valuable tool for refining dose extrapolation and improving the prediction of potential adverse effects.
• Further research with expanded datasets is warranted to fully leverage the potential of these modeling approaches.