The ADME TF aims to develop in silico skin penetration models using in vitro human skin penetration data for 25 chemicals solubilized in water. Since there are widely differing opinions on different in silico models, we evaluated 3 open source (DermWin™, CDC and the University of Surrey models) and 3 commercial models (DSkin, SimCyp and TCAT). Simulation of cutaneous distribution used a finite dose exposure scenario.
None of the models adequately predicted the amount of chemical that evaporated. This was important since the prediction of dermal delivery (DD) was improved when the evaporated amount was accounted for in the simulations. The ability to predict the amounts in skin layers varied between models, and the DD were generally over-predicted by 4 models and under-predicted by 1 model. Interestingly, measured partition and diffusion coefficients for the stratum corneum (SC) improved the DD prediction by one model from R2 = 0.23 to R2 = 0.57, while DD was predicted well by other models using QSAR values with higher R2 = 0.80. The effect of ethanol on DD was indicated by 5 of the models used, although they all over-predicted the DD.
In conclusion, our evaluation highlighted important differences in 6 models and identified required improvement of in silico models.
By Sébastien Grégoire, Ian Sorrell, Daniela Lange, Abdulkarim Najjar, Andreas Schepky, Corie Ellison, John Troutman, Eric Fabian, Hélène Duplan, Camille Genies, Carine Jacques-Jamin, Martin Klaric, Nicola J. Hewitt