Acute systemic toxicity testing, which is broadly utilized in pharmaceutical industrials and environmental protection agencies, is conducted to determine the relative health hazard of various chemicals. The traditional approaches for acute toxicity testing are costly, time-consuming, and require the use of many testing animals. Cell based assays, such as cytotoxicity testing, have been used as possible alternatives to animal acute testing. However, earlier studies have found poor correlation between these bioassay data and animal acute toxicity. To develop a predictive acute toxicity model, in this project we posit to construct diverse biological profiles of chemicals of environmental interest using large public toxicity bioassay data pools. To this end, we will implement an automated data mining method to explore the PubChem bioassay database. We have succeeded in preliminary studies to create a novel Quantitative Structure In vitro-In vivo Relationship (QSIIR) modeling workflow. In this project, we expect to apply this strategy to develop predictive animal acute toxicity models by employing the bioassay profiles for the modeling set compounds as extra biological descriptors. Moreover, we want to reveal the associated toxicity mechanisms by recognizing specific chemical-bioassay features in the modeling process. We expect to use the resulting model to prioritize toxic compounds of environmental interest for future animal toxicity studies.
Many drug candidates are eliminated as pharmaceutical leads due to unsuitable toxicity established in the course of the drug development process or in clinical studies. The cheminformatics animal toxicity predictor developed in this project is expected to directly evaluate chemical toxicity potential. In addition, the chemical-biological profiles for chemicals of environmental and pharmaceutical interest generated in this study could be used for in future studies of other complex toxicity endpoints.
Showing the most recent 10 out of 33 publications
