Identification of toxifying and detoxifying moieties for mutagenicity prediction by priority assessment

The search for structural subunits that affect compound toxicity cannot be manually performed on large databases. In addition, the a priori definition of important groups is impossible. Structural diversity requires the analysis of the complete data space and the selection of the details there present. A single substructure cannot be considered sufficient when assigning compound toxicity. In contrast, if we consider all the substructures in the database as the elements of a complete collection and if we can build a working hierarchy, the identification of the best feasible result using the available data is possible. If the database includes several significant examples, the results will be valuable. The use of a fragment-based description of a mutagenicity database together with the realization of a general hierarchy allows for the identification of the moieties that control the toxifying/detoxifying action of each compound. © 2011 American Chemical Society.