Adverse Outcome Pathway Network Guided High-Dimensional Modeling for Risk Assessment Regarding Drug Induced Liver Injury

• Activity Number: 304 - Risk Applications for Disease, Toxicology, and Biomarker Modeling
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Section on Risk Analysis
• Author(s): Dong Wang* and Kapil Khadka
• Companies: FDA National Center for Toxicological Research (NCTR) and National Center for Toxicological Research/FDA
• Keywords: expert opinion; high dimensional modeling; toxicology; dimension reduction; toxicogenomics

Abstract

There is heightened interest in incorporating high throughput assays into the evaluation framework for the risk of drug induced liver injury (DILI). However, the diverse and high dimensional nature of these data poses serious challenges for statistical modeling. We propose to integrate high throughput assay information with mechanistic knowledge in the form of expert opinion and literature findings to utilize the power of both mechanistic understanding and new technologies. We discuss our study using adverse outcome pathway (AOP) networks to provide a base model for incorporating high throughput data from L1000, CMap, and Tox21 for gene expression changes and nuclear receptor binding. A high dimensional learning model was then built to infer liver toxicity from this diverse array of molecular predictors. Our result suggests that current knowledge encoded in AOPs can be successfully utilized for dimension reduction for high throughput data and leading to capable predictive models. With continued improvement in AOP development and new testing technologies, combining mechanistic insight with high throughput data holds great promise in advancing DILI risk assessment.