You are here:
Development of an acute oral toxicity dataset to facilitate assessment of existing QSARs and development of new models
Citation:
Fitzpatrick, J., A. Karmaus, P. Pradeep, AND G. Patlewicz. Development of an acute oral toxicity dataset to facilitate assessment of existing QSARs and development of new models. ASCCT, Gaithersburg,MD, September 21 - 22, 2017.
Impact/Purpose:
Poster presentation at the ASCCT meeting on Investigating the feasibility of developing in silico models for acute toxicity
Description:
Assessment of the acute toxic potential of a substance is necessary to determine the adverse effects that might occur following accidental or deliberate short-term exposure. There are no accepted in vitro approaches available and few in silico models. Until recently, there had been a paucity of experimental in vivo acute toxicity data available for model development and evaluation. Here a large dataset of acute oral toxicity was compiled from different sources including the Hazardous Substances Databank (HSDB). Many of the studies were found to be limit tests, which report a LD50 as above a threshold typically 2000 mg/kg or 5000 mg/kg. This can present challenges for model development. A two-step approach was developed to model acute oral toxicity. In the first step, a random forest model was built to predict which substances would be above and below a LD50 of 5000 mg/kg. This model was constructed with a training set of 5931 substances with experimental LD50 values and ToxCast/Tox21 hitcalls as descriptors. On a test set of 1482 substances, the balanced accuracy of the model was 76% and its negative predictive value was 84%. In the second step, a ridge regression model was derived using a training set of 4164 substances with experimental LD50 values and ToxCast/Tox21 hitcalls as descriptors. For a test set of 1387 substances, 85% of the predictions made were within 1 log unit of their experimentally reported LD50 value. These models show considerable promise in predicting acute oral toxicity. This abstract does not reflect EPA policy.
URLs/Downloads:
ABSTRACT FOR ASCCT_FITZPATRICK_290617_GP_JC_FINAL.PDF (PDF, NA pp, 125.27 KB, about PDF)ACCTFITZPATRICK ACUTEORALTOXPREDICTIONS_POSTER_190917_FINAL.PDF (PDF, NA pp, 559.293 KB, about PDF)