Classification and prediction of toxicity of chemicals using an automated phenotypic profiling of Caenorhabditis elegans

Shan Gao, Weiyang Chen, Yingxin Zeng, Haiming Jing, Nan Zhang, Matthew Flavel, Markandeya Jois, Jing-Dong J. Han, Bo Xian, Guojun Li

Traditional toxicological studies have relied heavily on various animal models to understand the effect of various compounds in a biological context. Considering the great cost, complexity and time involved in experiments using higher order organisms. Researchers have been exploring alternative models that avoid these disadvantages. One example of such a model is the nematode Caenorhabditis elegans. There are some advantages of C. elegans, such as small size, short life cycle, well defined genome, ease of maintenance and efficient reproduction. As these benefits allow large scale studies to be initiated with relative ease, the problem of how to efficiently capture, organize and analyze the resulting large volumes of data must be addressed. We have developed a new method for quantitative screening of chemicals using C. elegans. 33 features were identified for each chemical treatment. The compounds with different toxicities were shown to alter the phenotypes of C. elegans in distinct and detectable patterns. We found that phenotypic profiling revealed conserved functions to classify and predict the toxicity of different chemicals. Our results demonstrate the power of phenotypic profiling in C. elegans under different chemical environments.