Development of Structure-Activity Relationships for Hepatotoxicity Using Proprietary Data.pdf

In this work we investigate the development of a structural alert based approach to the prediction of hepatotoxicity by using a proprietary data set of 731 chemicals and their associated sub-acute study data. The outcome of this research has demonstrated that given a suitable training set, new structural alerts for the prediction of hepatotoxicity can be derived using a clustering and knowledge based approach. The structural alerts as implemented do not compromise the confidentiality of the original data set. Such distilled knowledge can now be used as a supporting tool in the risk assessment of new chemicals.

Presented by Diana Suarez at Toxicology & Risk Assessment 2010 (TRAC), Cincinnati, Ohio, USA; 26^th - 29^th April 2010.