Integrated decision strategies for skin sensitization hazard (pages 1150–1162)
Judy Strickland, Qingda Zang, Nicole Kleinstreuer, Michael Paris, David M. Lehmann, Neepa Choksi, Joanna Matheson, Abigail Jacobs, Anna Lowit, David Allen and Warren Casey
Version of Record online: 6 FEB 2016 | DOI: 10.1002/jat.3281
The Interagency Coordinating Committee for the Validation of Alternative Methods (ICCVAM) evaluated a non-animal decision strategies to predict skin sensitization. Machine learning approaches integrated in vitro, in chemico and in silico data and six physicochemical properties for 120 substances to predict murine local lymph node assay (LLNA) outcomes. The seven models with the highest accuracy used a support vector machine with different combinations of predictor variables. The models outperformed individual non-animal methods and test batteries. This suggests that computational approaches are promising tools to effectively integrate data to identify potential skin sensitizers without animal testing.