Designing safer drugs: (Q)SAR-based identification of mutagens and carcinogens.

Abstract

Mutagenicity and carcinogenicity are chronic effects of primary concern for human health. A unifying approach to their mechanistic understanding is the recognition that many chemicals provoke both effects by electrophilic attack to the biological macromolecules, as such or after metabolism (genotoxic carcinogenicity). QSARs of individual classes of genotoxic carcinogens have contributed to the elucidation of the chemical determinants of this activity. Little work has been done on the epigenetic carcinogens, acting through non-genotoxic, very specific mechanisms. However, the existing QSARs for individual chemical classes are too few to be of real usefulness in the screening of masses of candidate drugs. Models for predicting the carcinogenicity of "any type" of chemicals have been proposed: prospective prediction exercises pointed to the serious limitations of most of these approaches. The best alternative is provided by panels of human experts. The above prediction exercises considered samples of general chemicals, thus we specifically addressed in this paper the issue of pharmaceutical drugs. We applied our expert knowledge to a database of drugs whose carcinogenicity/noncarcinogenicity status was known. Whereas most of the noncarcinogens were correctly identified, our prediction of carcinogens was less successful than with the general chemicals. Several carcinogenic drugs did not show recognized structural alerts, and supposedly acted by epigenetic mechanisms. Whereas the contribution of human experts is highly valuable in this phase (e.g. priority setting), more work is necessary on: a) epigenetic carcinogens; b) efficient computerized models.