Using a genetic algorithm to identify common structural features in sets of ligands

This article describes a program for pharmacophore mapping, called MPHIL (Mapping Pharmacophores in Ligands). Given as input a set of molecules that exhibit some common biological activity, MPHIL identifies the smallest 3D pattern of pharmacophore points that has at least m (a user-defined parameter) points in common with each of the input molecules. The program thus differs from existing programs for pharmacophore mapping in that it does not require all of the molecules to share exactly the same pattern of points, although it will find such a common pattern if it does, indeed, exist. MPHIL uses a genetic algorithm (GA) approach in which an initial, and very rapid, GA is used to suggest possible combinations of points that are then processed by the second GA to yield the final 3D pattern.