Structure-activity maps for visualizing the graph variables arising in drug design.

Abstract

Structure-activity problems are characterized by the topological and topographical character of the structural information determining the activity. Traditional statistical methodology requires that this predictive information be mapped to a vector space. To circumvent this vexing conversion of structural information to vector form, the edge-deletion metric is defined on the space of chemical graphs that defines the topology of the molecules. This paper proposes structure-activity maps and transformation-effect maps for directly visualizing the structure-activity relationships. The maps are illustrated using the hypotensive activities of clonidine analogs and the sweet taste of Perillartine analogs.