Abstract

Introduction: Sustained activation of AhR is the molecular initiating event (MIE) in several adverse outcome pathways (AOPs)1 :

• Lethality to fish and bird embryos
• Causing uroporphyria in birds
• Producing liver tumors in rodents

Results from a quantitative high throughput screening (qHTS) assay, run as part of the U.S. Tox21 program2 designed to detect AhR activators, were recently made public. 3 Our objective was to use those data to develop an in silico classifier to identify AhR activators from their structure.

• Compounds with AC50 < 100 µM were categorized as activators (“positives”).
• A few of the most potent activators are below (SID = PubChem substance ID).
• The data set is highly skewed towards negatives, with only 11% of the compounds tested being categorized as positives.

Preparation of the Modeling Set: AID743122 (AID = PubChem bioassay ID) has two components: a cell-based AhR activation assay (AID743085) run in triplicate at 15 concentrations, and a cell viability counter screen (AID743086). AhR results were reported in binary format: 1 for positives and 0 for negatives. ADMET Predictor™ 4 descriptors are parameterized for H, B, C, N, O, P, S, Cl, Br, I, and F, so compounds containing other elements were removed. Ionic salts were also removed. Salts and other mixtures were resolved based on molecular size. Categorization of molecules tested more than once was by majority vote, with ties being set aside. The data set was then divided into a training set for developing models and a completely external (blind) test set for validating them after models were trained. Both subsets contain a similar balance of positives (about 11%) to negatives (about 89%).

SETAC North America 36th Annual Meeting, November 1-5, 2015, Salt Lake City, UT

Michael S. Lawless, Jayeeta Ghosh, and Robert D. Clark