Description:
To make a comprehensive assessment of three different multi-parameter in silico absorption models, i.e., GastroPlusTM, Intellipharm&[reg]PK and a microscopic mass balance model, based on the work of Oh et al. 1993 (Pharm. Res. 10:264-270).46 literature compounds, with a wide range of physico-chemical and biopharmaceutical properties were selected. Simulation input parameters such as solubility and permeability were measured, whereas other input data (e.g. dose, log D, pKa, Mw) were taken from the literature. In an attempt to assess the importance of media chosen for the solubility measurements, the solubility of all drugs were measured in buffers, simulated GI fluids (FaSSIF) and actual human GI fluids (HIF). The fraction dose absorbed (Fabs) was predicted using all models and a comparison was made with actual human data.When comparing the predicted fraction dose absorbed to published human data, there was an obvious correlation (r2&[approx]0.6) for all models but also a few clear outliers. Of the 3 major outliers, two drugs showed significant levels of efflux in the Caco-2 cell assay and one drug showed signs of active transport. If effluxed or actively transported compounds were excluded, the correlations were significantly improved and good correlations were obtained (r2&[approx]0.7-0.8). The choice of media for the solubility measurements had no major impact on the relationships between predicted and observed human Fabs, although a close examination of solubility data in isolation showed the propensity for buffer to under-estimate and FaSSIF to over-estimate HIF solubility for poorly soluble compounds.In this study we have shown that all three tested in silico models can be used to predict the human fraction dose absorbed of passively absorbed compounds with high confidence. Hence, the models are very useful tools in assessing the developability of compounds early in the drug discovery process. |