Predicting DILI liability by transcription factor profiling

Period of Performance: 08/01/2017 - 07/31/2018


Phase 2 SBIR

Recipient Firm

Attagene, Inc.
Principal Investigator


PROJECT SUMMARY Drug-induced liver injury (DILI) is the main reason for drug attrition during development and a leading cause of post-market drug withdrawal. Here, we propose a systems biology approach to detect drug candidates with DILI liabilities at early stages of development. This approach is based on the assessment of drug-induced perturbations of multiple signal transduction pathways in hepatocytic cells. For that, we use Attagene multiplexed reporter technology, the FACTORIAL?,that enables quantitative assessment of the activity of multiple transcription factors (TFs), proteins that regulate gene transcription. The FACTORIAL has been extensively validated by screening thousands of environmental toxicants for the U.S. EPA ToxCast project. Through this effort, we discovered specific ?TF signatures? for many classes of biological activities. In preliminary studies, we evaluated TF signatures for a small panel of drugs with DILI liabilities and found a common pattern. Within certain concentration range, drugs' TF signatures reflected their primary activities. However, at some inflection points (COFF), these signatures transformed into distinct, off-target, signatures. We found common off-target TF signatures shared by different classes of DILI drugs and identified underlying mechanisms for some of those common TF signatures, including mitochondrial malfunction, DNA damage, and lipid peroxidation. Based on these findings, we developed a simple model wherein DILI mechanism is inferred from the off-target TF signature, whilst DILI probability is defined by the CMAX/COFF ratio, where CMAX is the maximal therapeutic drug concentration. Most remarkably, our data suggest the feasibility of using this model to predict idiosyncratic DILI, the task unattainable with existing technologies. The overarching objective of this proposal is to establish TF profiling as a tool for DILI prediction. To do that, we will obtain TF signatures of a collection of 396 drugs classified by the FDA as DILI and no-DILI concern drugs. These signatures will be used as a training set. We will identify clusters of common DILI-specific off-target TF signatures and annotate the underlying biological activities, using ATTAGENE DB of reference TF signatures. To validate the off-target TF signatures as potential bioactivity markers, we will compare these with data by functional assays for known DILI mechanisms. Furthermore, we will determine the predictive value for the CMAX/COFF parameter for stratifying DILI from non-DILI drugs. The predictive values of obtained DILI-specific TF signatures and the CMAX/COFF parameter will be optimized using a validation set of exhaustively characterized in functional assays drug candidates, provided by pharmaceutical industry and DILI-sim consortia.