FDA has identified "Modernizing Toxicology to Enhance Product Safety" as one of its 8 priority areas in its Regulatory Science Strategic Plan. The plan seeks to develop new approaches to better understand how drugs lead to adverse effects. One of the approaches of interest is the use of “systems pharmacology”, an emerging scientific discipline that blends biology with computational modeling to untangle the intricate networks of interactions between genes, proteins, metabolites, and other small molecules within cells. This approach also forms the basis for the Agency’s mechanism-based drug safety assessment and prediction program, "Pharmacological Mechanism-Based Drug Safety Assessment and Prediction", launched in 2011.

The PredicTox pilot program brings together multiple stakeholder groups (FDA, NIH, patient groups, drug manufacturers, and academia) to leverage collective knowledge, technical expertise, data, funding, and other resources to systematically explore systems pharmacology approaches to better understand adverse events (AEs).

While systems pharmacology approaches can be applied to the development of predictive models for any class of drugs or AE, the PredicTox pilot seeks to first provide a “proof of concept”, focusing on drugs and biologics (tyrosine kinase inhibitors [TKIs] and monoclonal antibodies [mAbs]) that target tyrosine kinases and one class of toxicity, cardiac AEs . Targeted therapies are a rapidly growing strategy for oncology treatment, but they are not without AEs, including cardiac AEs, an area of intense importance for patients, the FDA, and pharmaceutical manufacturers. Learnings from the PredicTox pilot can then be applied to other drug classes and/or other toxicities to improve drug safety assessment. The systems approach also has the potential, in future years, to bolster pharmacovigilance efforts by providing mechanistic/biological insight to help evaluate potential safety signals. In addition to providing evidence to support or refute signals, the mechanistic knowledge provided by this approach can be used to generate hypotheses for prospective investigation in the search for safety signals.

The PredicTox pilot will be conducted as an iterative, phased project over the course of several years. This proposal focuses on Phase I of the pilot, which will build and populate a data sharing platform. Phase I will provide near-term scientific value; build a public-private partnership; leverage resources and participation from various sources (FDA, NIH, patient groups, foundations, and drug manufacturers); and assess the long term program feasibility of using a systems approach to better understand AEs. Phase I will assemble the data and partners necessary for future phases involving the development of systems pharmacology based predictive models. FDA cannot build these mechanism-based, predictive approaches alone. Multiple sectors must partner in order to bring together necessary resources---including data, expertise, and funding--- for building this capacity. PredicTox goes beyond previous and current individual company, academic, or FDA efforts, by bringing together in one place:

• A breadth of mechanisms of toxicity across this class of drugs/biologics
• Data beyond what is included in a regulatory submission
• A systematic vocabulary and computational framework (ontology) for data importation and integration
• An integrated software platform--- a significant need in the field of systems pharmacology--- to enable mining and analysis of very different types of data not typically combined
• A data warehouse containing both publically available data and cross-company data for researchers to analyze collectively

The long term goal is for PredicTox to move from data mining to development of computational models capable of linking events at the molecular level with events at the clinical level (AEs), to create improved safety assessment tools including:

• Development of safer drugs with less cardiotoxic potential
• Treatment strategies to mitigate potential cardiac toxicity
• Nonclinical screens and models for drug safety assessment to identify potential problems early on in the drug development pipeline - faster, cheaper drug development
• Clinical diagnostics for detection and early intervention
• Clinical risk prediction based on precision medicine (e.g. individual susceptibility)

Much of the scientific knowledge, technological capacity, and data required to better understand TKI-associated cardiac AEs already exist, but it has not been harnessed in a systematic way across sectors and disciplines. Key to this project is the ability to obtain data from publicly available sources as well as from pharmaceutical partners. Data collection includes:

• Nonclinical data

• Molecular and cellular data including kinase binding panels and -omics

• in vivo cardiovascular data, including imaging parameters

• standard in vitro and in vivo pharm/tox data package

• Clinical Trial data

• Cardiac AE, medical history and cardiac risk factor information

• Clinical Pharmacology (pharmacokinetic and metabolic parameters)

• Cardiac Imaging data

• Blood biomarkers (cardiac troponin, NT-pro-BNP)

Additionally, FDA, NIH, industry and academic partners are providing active scientific contributions and leadership

• Predictive informatics, Computational Biology, Chemoinformatics, Toxicoinformatics
• Nonclinical, Translational Biology and Clinical safety expertise
• Kinase Biology and Medicinal Chemistry expertise

A large part of the preliminary project work will focus on the coordination and agreement required for all of the partners to be able to work together. This includes, but is not limited to, reaching agreement on issues such as:

Legal/Partnership structure:

• Which TKIs/mAbs are included?
• What data will be shared for each TKI/mAb
• Who may access the database?
• Are there limitations on the type of research that can be conducted?
• Will those conducting research be required to report findings to the PredicTox Scientific Steering Committee, share findings with other data contributors, or share findings externally?

Technical Parameters:

• A common language and framework (ontology) for the agreed upon data elements
• Data gathering, formatting, and curation
• Selection of data platform vendor
• ability to integrate pre-clinical, clinical, molecular, genetic, and -omics data
• data mining tools and software included/available
• compatibility with common data mining and visualization software
• scalability and flexibility to create less/more functionality
• compatibility with each partner’s existing data format
• Design, development, and management of data platform
• Security of the platform

• A data sharing platform for nonclinical and clinical data that will expand knowledge of:
• biological pathways/networks underlying TKI/mAb-related cardiac AEs
• incidence, severity, and nature of cardiac AEs as well as patient-specific risk factors
• Inventory of data and identification of data needs for future modeling phases
• Systematic framework (ontology) for knowledge management, data Integration and decision support that will provide the necessary infrastructure for linking AEs with underlying biological mechanism
• A legal framework and governance structure for sharing pre-competitive drug safety data
• A valuable resource for FDA, academia, and industry to mine in the development of biomarkers, nonclinical screens, clinical screens, and predictive models in future phases