You are here:
QSAR Modeling of Caco-2 Permeability for the Estimation of Oral Bioavailability
Citation:
Honda, G., R. Sayre, C. Strock, D. Angus, R. Dinallo, R. Pearce, R. Thomas, AND J. Wambaugh. QSAR Modeling of Caco-2 Permeability for the Estimation of Oral Bioavailability. Presented at Society of Toxicology Annual Meeting, Baltimore, MD, March 10 - 14, 2019. https://doi.org/10.23645/epacomptox.7841102
Impact/Purpose:
This is a poster for presentation at the Society of Toxicology annual meeting. The research concerns improving high throughput modeling of oral exposure to chemicals such that bioactivity:exposure ratios can be calculated more accurately.
Description:
Toxicokinetics can be used to extrapolate an oral equivalent dose from in vitro bioactivity data for comparison with potential external exposure rates, thereby providing an estimate of risk. To more accurately predict the oral equivalent dose, it is desirable to estimate the oral bioavailability (Fbio), that is, the fraction of the oral dose that is actually available to the body. Caco-2 assays inform estimates for Fbio by providing a measure of the in vitro apparent permeability (Papp, 10-6 cm/s) across a membrane of human colon carcinoma cells. This permeability is highly correlated with the fraction of chemical absorbed (Fa) in the gut and the effective permeability rate (Peff, 10-4 cm/s) through the epithelium of the small intestine (Artursson, et al. 2001). Predicted P¬eff may then be used with in vitro measured intrinsic hepatic clearance (Clint) to estimate the fraction of chemical surviving gut metabolism (Fg) (Yang, et al. 2007). Subsequently, Fbio can then be determined by combining Fa and Fg with the fraction of chemical surviving first pass hepatic clearance, with the latter predicted using Clint and the fraction unbound in plasma. In this work, we developed a random forest QSAR model to predict Papp using recently measured values for environmental chemicals and literature values for pharmaceuticals. We then used the predicted Papp to estimate Peff, Fa, Fg, and Fbio for comparison to literature values. Using a training set of Papp for 160 chemicals, the QSAR model provided reasonable prediction of the Papp for the validation dataset of 315 chemicals, yielding an RMSE of 0.60 for the log10 transformed values. Using the QSAR predicted Papp to predict Peff gave an RMSE of 0.69 for the log10 transformed values. Predictions for Fa in humans made using the model reported by Darwich et al. (2010) had an RMSE of 0.26. Estimates for Fg in humans using the Qgut model (Yang et al. 2007) gave an RMSE of 0.26. Subsequent estimates of Fbio for human and rat had RMSE’s of 0.32 and 0.47 respectively. With additional open-source models to predict Clint and fup, it would be possible to make predictions for Fbio entirely using in silico methods. This abstract does not necessarily reflect U.S. EPA policy. Artursson, et al. Adv. Drug Deliv. Rev. 2001, 46, 27-43. Yang, et al. Current Drug Metab. 2007, 8, 676-684.
URLs/Downloads:
DOI: QSAR Modeling of Caco-2 Permeability for the Estimation of Oral Bioavailability
HONDA_CACO2_SOT2019_FINAL.PDF (PDF, NA pp, 2625.979 KB, about PDF)