The results of predictive toxicogenomics investigations over the past 6 years reviewed in this report have shed new light on the potential of molecular expression analysis to more properly classify both genotoxic and nongenotoxic carcinogens and to predict the carcinogenicity of untested chemicals. Predictive toxicogenomics uses global molecular expression data resulting from genomic perturbation (e.g., transcription or gene expression profiles) to predict a toxicological outcome, such as carcinogenicity. The classification of carcinogens has become an essential and highly debatable component of cancer risk assessment largely because of the default assumptions that drive regulatory decision-making regarding the presumed linearity of the dose-response curve for genotoxic carcinogens. Nongenotoxic mechanisms of carcinogenesis complicate the well-established relationship between genotoxicity and carcinogenicity and challenge the interpretation of the results of rodent carcinogenicity studies in terms of their relevance to humans. Although the number of presumed nongenotoxic rodent carcinogens has dramatically increased over the past two decades, the fact remains that more than 90% of the known human carcinogens are detected in conventional short-term tests for genotoxicity and induce tumors at multiple sites in rodents. In toxicogenomics studies, a strong DNA damage response at the gene expression level suggests direct DNA modification whereas increased expression of genes involved in cell cycle progression is more characteristic of the indirect-acting agents such as those that induce oxidative stress. Metabolism genes are prominently represented among gene expression profiles that discriminate nongenotoxic modes of action (e.g., cytotoxicity and regenerative proliferation, xenobiotic receptor agonists, peroxisome proliferator-activated receptors, or hormonal-mediated processes). The evidence accumulated to date suggests that gene expression profiles reflect underlying modes or mechanisms of action, such that they will be useful in the prediction of chemical carcinogenicity, especially in conjunction with conventional short-term tests for gene mutation, chromosomal aberration and aneuploidy.