ABSTRACT
The application of genomics techniques to the field of toxicology has created a new science subdiscipline, toxicogenomics. Toxicogenomics is the study of the structure and transcriptional output of the entire genome as it relates and responds to adverse xenobiotic exposure. This chapter focuses on one study that applied unsupervised clustering to expression profiles from rats treated with 15 different known hepatotoxins: allyl alcohol, amiodarone, Aroclor 1254, arsenic, carbamazepine, carbon tetrachloride, diethylnitrosamine, dimethylformamide, diquat, etoposide, indomethacin, methapyrilene, methotrexate, monocrotaline, and 3-methylcholanthrene. These agents have been shown to cause a variety of toxic hepatic effects, including hepatocellular hypertrophy, DNA damage, necrosis, steatosis, and cholestasis, among others. The difference between supervised and unsupervised learning is that with supervised learning, the user has previous knowledge from outside the microarray experiment concerning the output value or the input variable. This information is used to direct the analysis of the microarray experimental results.
