Abstract

Accurate Prediction of Organ-Specific Toxicities

Daniele Zink, Principal Research Scientist and Team Leader, Institute of Bioengineering and Nanotechology, Agency for Science Technology and Research (A*STAR)

Predicting toxicity to internal organs with alternative methods is currently a major challenge. We have developed the first methods that predict nephrotoxicity in humans with high accuracy. These methods include a high-throughput platform, which combines high-throughput screening (HTS) with phenotypic profiling and machine learning. With this technology toxicity to renal proximal tubular cells (PTC) could be predicted with 82% or 89% test balanced accuracy, depending on whether primary human PTC or a human PTC line were used. Based on the same technology we have developed a HTS platform for the prediction of hepatotoxicity in humans. This HTS platform has been validated with ~90 compounds that were separated into training and test sets, and both test sensitivity and specificity are in the range of 80%. In addition, we have developed a HTS platform for the prediction of vascular toxicity. Our predictive HTS methods are currently combined with our predictive approaches based on human induced pluripotent stem cells.
 
While the HTS platforms are robust, suitable for rapid compound screening and produce binary (yes/no) results with respect to toxicity prediction, we develop complementary microphysiological systems (MPS) for repeated dose testing. Results obtained with a kidney-specific MPS show a high correlation between IC50 values obtained in vitro and the lowest toxic doses in humans. It is expected that respective MPS can greatly facilitate in vitro to in vivo extrapolations (IVIVE). Current difficulties in this area are one of the major obstacles that prevent applications of in vitro methods in risk assessment.