Online Program Home
My Program

Abstract Details

Activity Number: 287
Type: Invited
Date/Time: Tuesday, August 2, 2016 : 8:30 AM to 10:20 AM
Sponsor: Section on Risk Analysis
Abstract #318241 View Presentation
Title: Predicting Chemical Dose-Response Toxicity Through Chemical Structure Activity Relationships
Author(s): Matthew Wheeler*
Companies: CDC/NIOSH
Keywords: Machine Learning ; Functional Data Analysis ; Baysian Nonparametrics
Abstract:

Quantitative structure activity relationship (QSAR) models link physical chemical outcomes to endpoints related to toxicity. Traditionally, these endpoints are scalar values that are regressed upon using linear models, treed regression, or neural nets. For example, there is a large literature developing regression models to that link chemical properties to the LD50. Though such endpoints give some idea of toxicity, they give no indication of the relative potency of the chemical across a spectrum of doses; ideally, one would be interested in methodologies that link chemical properties to dose-response relationships. This talk investigates one such QSAR model that is used to model bioassays from the US EPA's ToxCast database. The model is presented along with results that show accurate predicted dose-response curves from a large hold out dataset.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association