NIEHS grantees developed a new algorithm to predict the toxicity of new and untested chemicals by evaluating their similarity to chemicals that have been previously tested for toxicity. The algorithm compares chemical components from tested compounds with those of untested compounds. Using mathematical methods, the algorithm identifies similarities and differences in the chemical structures that may factor into toxicity.
Low-cost, cell-based tools for toxicity testing have potential as alternatives to animal testing. However, incorporating cell-based assays into chemical toxicity evaluations requires significant data curation and analysis. To address this problem, the researchers developed a computational method that extracts useful data from cell-based assays and predicts animal toxicity based on relevant toxicity mechanisms.
To fine-tune the algorithm, the researchers began with 7,385 compounds for which rat acute oral toxicity data are known. They compared the data with other data on the same chemicals in PubChem, a National Institutes of Health database on millions of compounds. They then tested the algorithm on 600 new compounds.
For several groups of chemicals, the algorithm had a 62-100% success rate in predicting levels of oral toxicity. By comparing relationships between sets of chemicals, they also discovered new factors that can determine a chemical’s toxicity. According to the authors, this data-driven model provides a tool to prioritize chemicals that may need more comprehensive testing in animals before use in commerce.
Citation: Russo DP, Strickland J, Karmaus AL, Wang W, Shende S, Hartung T, Aleksunes LM, Zhu H. 2019. Nonanimal models for acute toxicity evaluations: applying data-driven profiling and read-across. Environ Health Perspect 127(4):047001.
Read more at: https://factor.niehs.nih.gov/2019/6/papers/dert