Adverse drug events are a major challenge in drug development and clinical practice, partly due to the limitations of preclinical assays that poorly mimic dynamic human exposures. Here, we combine physiologically based pharmacokinetic (PBPK) simulations with primary human liver spheroids to reproduce clinically realistic drug concentration–time profiles and enable mechanistic assessment of hepatotoxicity.
Ten hepatotoxic drugs were tested under therapeutic and toxic exposure regimens. Time-resolved transcriptomics, proteomics, targeted metabolite analysis, and viability measurements were collected over 14 days and integrated with context-specific metabolic models and probabilistic flux simulations. Functional enrichment revealed consistent alterations in cholesterol and central carbon and nitrogen metabolism. Metabolite measurements showed that toxic exposure led to increased catabolism of amino acids. Further metabolic modelling showed shifts away from glycolysis toward fatty acid, bile acid, and amino acid metabolism. Model complexity increased over time, with expanding reaction networks reflecting adaptive metabolic responses to stress. A detailed case study of isoniazid is presented to illustrate the integration workflow. Accounting for alternative optima and using z-scores improved interpretability. Proteomics and transcriptomics correlated moderately (r=0.32), but proteomics aligned with flux predictions (r=0.19), particularly for secondary pathways such as fatty acid oxidation, emphasizing the value of quantitative metabolite exchange constraints to capture central metabolism. This framework bridges pharmacokinetics and metabolic flux analysis to elucidate mechanisms of drug-induced liver injury, advancing translational safety assessment and therapy design.

Join Zoom Meeting

Please register in advance for this meeting:

univienna.zoom.us/meeting/register/uSH_PvW0Qu6XW6UAo3eTsg

After registering, you will receive a confirmation email with information on how to join the meeting.