Precipitation as key parameter for early oral drug absorption predictions, Part II: Development of a predictive model

Abstract

Accurate absorption predictions for new chemical entities (NCEs) are key in drug discovery, as they enable the early identification of compounds with suboptimal absorption characteristics. Common methodologies, such as the biopharmaceutical classification system (BCS), often lack quantitative insights, or they require extensive datasets at early stages, as seen in Physiologically Based Biopharmaceutics Modeling (PBBM). Furthermore, existing tools frequently overlook the impact of precipitation of weakly basic drugs on human absorption. To bridge this gap, we developed a novel oral absorption prediction tool that uses easily obtainable in vitro assay results while systematically investigating the relationship between in vitro precipitation and in vivo absorption. The underlying dataset, which is compiled in Part I of this study, includes in vivo human absorption data alongside in vitro permeability, solubility, and precipitation data for 17 basic compounds. In this part of the study, the dataset underwent a comprehensive statistical analysis. While regression analysis proved inadequate, a decision tree using the Classification and Regression Trees (CART) algorithm was constructed. The assessments from this algorithm led to the inclusion of both precipitation and permeability in the final decision tree, while solubility was excluded. This finding underscores the critical role and added value of in vitro precipitation compared to solubility. This intuitive tool is particularly well-suited for drug discovery, as it facilitates ranking and selection of drug candidates based on their oral absorption characteristics. Additionally, it can be easily adapted by other researchers, allowing for the incorporation of various in vitro assays and larger datasets to meet specific research needs. This flexibility paves the way for future advancements in the field.