The aim of the PredicTox Knowledge Environment (KE) project is to advance the science of adverse event prediction by creating an integrated web-based knowledge environment consisting of multiple interconnected databases that capture quantitative multiscale biology of drug action. We refer to this as PredictTox-KE. The databases will contain information on clinical indications, preclinical physiological data, and cellular regulatory networks. Although many datasets relevant to drug-induced toxicities currently exist, the advantage of a centralized repository is that data can be normalized to ontologies to ensure consistent definitions. As such, all data in PredictTox-KE will be organized within a computable framework that enables integration and analysis.
Creation of the PredicTox-KE will enable an integrative approach to prediction of major adverse events associated with therapeutic agents. This will be done as a collaborative project between researchers at Icahn School of Medicine at Mount Sinai (ISMMS) and Critical Path Institute (C-Path). Data from academia, industry and publicly available databases will be integrated, and querying tools and analytical models will be developed to enable predictive toxicology. In this initial two-year demonstration project, we propose the creation of an initial proof of concept database containing clinical, cellular, and animal data related to adverse events (i.e. cardiomyopathy) caused by small-molecule (the “NIB” class; e.g. imatinib, sorafenib, sunitinib) or antibody (the “MAB” class; e.g. trastuzamab) protein kinase inhibitors used to treat cancer. Although these targeted therapeutics are being successfully used to treat several cancers, mechanisms underlying cardiotoxicity are poorly understood, which makes improved prediction of these serious adverse events an urgent issue.
The PredicTox-KE is expected to enable building of computational models for prediction of drug toxicity and, in the longer term, predictions for personalization of therapy to mitigate adverse events. The PredicTox-KE will store clinical, preclinical/physiological, cellular, and molecular (omics) data. In the longer term we hope to incorporate structural and genomic data and link to relevant genomic and epigenomic data in standard National Library of Medicine databases. We will harmonize data across domains and develop the databases and technical infrastructure to enable integrated analyses. Users of the PredicTox-KE will be able to build statistical, network and dynamical computational models to enable simulations that test hypotheses about drug actions and mechanisms underlying adverse events and prediction of adverse events.
Advance the science of adverse event prediction by creating computational models for prediction of drug toxicity and, in the longer term, predictions for personalization of therapy to mitigate adverse events.