The search for active substructures in structure-activity studies

Abstract

We consider the problem of location an active fragment (substructure) common to a class of biologically active compounds and presumed responsible for their biological activity (therapeutic or toxic). Our approach is graph-theoretical in that molecules are represented by suitable graph-theoretical invariants. Specially weighted paths in the molecular graph are adopted as descriptive elements. By selecting different sets of atoms one searches for a fragment that best represents the relative activities of the compounds. As an illustration we consider a dozen nitrosamine mutagens and analyze the cases of five-, six-, and sevenatom fragments. The approach clearly indicates that a specific seven-atom fragment (for molecules with up to 11 nonhydrogen atoms) can account for the relative mutagenic activities of the nitrosamines considered.