Improving In Silico Toxicity Predictions and Molecular Design Guidelines Through Mechanism-Based Modeling and Systems Analyses of High Throughput Data: a Case of Oxidative Stress

EPA Grant Number: FP917793
Title: Improving In Silico Toxicity Predictions and Molecular Design Guidelines Through Mechanism-Based Modeling and Systems Analyses of High Throughput Data: a Case of Oxidative Stress
Investigators: Melnikov, Fjodor
Institution: Yale University
EPA Project Officer: Lee, Sonja
Project Period: September 1, 2015 through August 31, 2018
Project Amount: $132,000
RFA: STAR Graduate Fellowships (2015) RFA Text |  Recipients Lists
Research Category: Academic Fellowships

Objective:

This aims of this project is two-fold. First, we will improve in silico tools to assess chemical toxicity and evaluate substitutes for substances likely to induce toxicity via oxidative stress (OS). A particular emphasis will be placed on identifying modes of action (MOA) and molecular initiating events (MIE) involved in OS production and response. Second, we aim to advance the applications of statistical learning algorithms in predictive toxicology by evaluating analytical methods that can effectively elucidating chemical toxicity adverse outcome pathways (AOPs) and interaction among toxic endpoints using large HTS data repositories.

Approach:

We will first build predictive toxicity models for OS-derived toxicity for specific MIEs; furthermore we will elucidate the AOPs through network analysis of HTS data and incorporate the insight into toxicity prediction. The quality of toxicity models relies largely on the quality of chemical descriptors and toxicity data used in model development. To assure highest reliability, toxicity models will be developed based on Tox21 and ToxCast databases that present the largest HTS data collection and curation efforts to date. Chemical properties to assess bioavailability, electrophilic reactivity, and specific non-covalent interactions will be evaluated. Examples of these include octanol-water distribution, and Abraham coefficients, dipoles moments, ionization potential, frontier orbital energies, and molecular shape parameters. All models will be externally validated with in vitro and in vivo data and delivered with an explicit applicability domain. Furthermore, in vivo validation will help establish relevance of in silico predictions to environmental and health effects in the real world.

Expected Results:

We expect to develop predictive and explanatory models for chemical toxicity through the mechanisms of oxidative stress. The models will be used to assess chemical toxicity, prioritize chemicals for further testing, and inform mechanisms of chemical toxicity. Furthermore the chemical insights from properties and insights of toxic substances will be used to design new alternatives for hazardous substances.

Supplemental Keywords:

Predictive Toxicology, Molecular Design, High Throughput Data, Alternatives Assessment, Oxidative Stress, in silico models

Progress and Final Reports:

  • 2016
  • 2017
  • Final