correlate the various patterns of altered gene expression caused by chemicals with unknown mechanisms with those changes known to be caused by toxicants acting via known mechanisms or by identified mutations in developmental processes. In this way, the effects of known and unknown compounds could be compared, as well as the relative strength of particular compounds in eliciting particular modes of action.
Due to expense and time, mammalian tests have a lower throughput of chemicals than described for information levels 1 and 2, perhaps 102 per year. (For reference, several hundred rat tests are now done per year in preparation for evaluations of human risk.) In the context of risk assessment (i.e., exposure pathways, human use, and environmental release), the tested chemicals would be those requiring a quantitative assessment of human risk. Chemicals that alter signaling pathways, molecular-stress pathways, or checkpoint pathways in assays at information levels 1 and 2 would be scrutinized at level 3 to ascertain in vivo mammalian effects.
Test animals would likely be the mouse and rat. At present, the mouse has the advantage of much greater available genetic information and ease of genetic manipulation, but the rat genome project will soon make that information available as well. Although it is widely recognized that mice and rats differ from humans in their metabolism of chemicals by DMEs and in their developmental responses to chemicals, these differences have not been well analyzed or characterized. Assays at level 3 will benefit greatly in the near future from research comparing mice and humans with respect to the uptake, conversion, and clearance of chemicals. The experimental means to analyze the metabolic differences are available, and the genetic means are available to reduce or eliminate the differences, thereby improving the accuracy of extrapolation across species. For example, mice could be provided with human transgenes to give a human-like profile of DMEs and other proteins of toxicokinetic importance. In cases in which differences cannot be eliminated, the differences should be well characterized so that bounds can be set on default corrections. Test animals can also be genetically modified to approximate the more sensitive rather than less sensitive members of the human population. Research is focusing on the analysis of human polymorphisms, and many of them could be approximated by genetically modified test animals. Although it can be argued that genetically modified animals might give false positives, the opposite can be argued—namely, that when a chemical has no negative effect in the sensitized test animal, the result would not require so many orders of magnitude of default correction when the extrapolation to humans is made at the time of risk assessment.
The molecular-stress and checkpoint pathways are expected to be activated by a wide range of toxicants and to show toxicant effects. Recently, these path-