Prediction of Chemical Carcinogenicity in Rodents by Machine Learning of Decision Trees and Rule Sets

Dennis Bahler and Douglas W. Bristol

We constructed toxicity models in the form of decision trees and rule sets by machine learning, using as a training set recent results from rodent carcinogenicity bioassays conducted by the National Toxicology Program (NTP) on 226 test agents. We performed 10-way cross-validation on each of these models to approximate their expected error rates on unseen data. We are using the models to offer prospective predictions of the carcinogenicity of the thirty test agents in the second phase of the NIEHS Predictive Toxicology Evaluation experiment (PTE-2) now underway.

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.