Towards a novel toxicity prediction pipeline, combining PBPK models and mechanistic cellular adversity models

There is a strong need for animal-free methods that properly predict drug adversity. Especially drug-induced liver injury (DILI) is a common adverse effect that is difficult to predict with current methodology. In this study, we developed a data-driven computational adversity model for predicting the relative number of apoptotic cells in vitro , based on stress pathway activity invoked by exposure of liver cells to nitrofurantoin, diclofenac, and ketoconazole. Next, we adapted this adversity model to represent an in vivo situation and developed physiologically-based pharmacokinetic (PBPK) models for all three drugs to simulate the concentration of the drugs in the liver of humans and rats. Coupling the PBPK, stress pathway and adversity models generated quantitative systems toxicology models for rats and humans at clinically-relevant doses. For all three drugs, low levels of toxicity were predicted in rats and humans, which is consistent with in vivo observations. A detailed analysis of modifying factors (sex, age) showed that the amount of apoptotic cells is expected to increase for females and elderly people compared to males. Coupling the different models generated a novel modelling pipeline that is an important step towards in silico DILI predictions.