Investigation of an in vitro hepatotoxicity dataset to search for co-occurrence of signals amongst assays.
We investigated whether biological fingerprints based on in vitro assay data could be used to look for significant correlations between appropriate in vitro assays and the prediction of in vivo hepatotoxicity. A dataset of compounds with relevant in vitro data was used to create biological fingerprints, where each bit represented the result of an in vitro test. These were analysed using the KNIME platform to identify where the predictive probability of a combination of in vitro results (AND logic) was better than an individual in vitro assay result, in correctly predicting a positive in vivo finding. Although the data is limited, our findings show that, some combinations of in vitro assays may offer better predictive potential than individual assays and also offer a mechanistic rationale for key events at a molecular level. This work supports findings reported in recent publications about how the interaction of a single compound with different subcellular structures such as the mitochondria and drug transporters may contribute to damage in hepatocytes.
Poster presented by Sebastien Guesne at FutureTox III, Arlington, USA; 20th November 2015.