ACTIVE SITE DRIVEN LIGAND DESIGN: AN EVOLUTIONARY APPROACH

Abstract

An evolutionary approach for designing a ligand molecule that can bind to the active site of a target protein is described in this article. An earlier attempt in this regard assumed a fixed tree structure of the ligand on both sides of the pharmacophore, and used a genetic algorithm for optimizing the van der Waals energy. However, it is evident that knowledge about the size of the tree is difficult to obtain an a priori. Moreover, it will also change from one active site to another. This limitation is overcome in the present article by using variable string length genetic algorithm (VGA) for evolving an appropriate arrangement of the basic functional units of the molecule to be designed, whose size may now vary. The crossover and mutation operators are appropriately redesigned in order to tackle the concept of variable length chromosomes. Once the geometry of the molecule is obtained, the possible three-dimensional structure and its docking energy is determined. Results are demonstrated for five different target proteins both numerically and pictorially. It is found that not only does the molecule designed using variable length representation, in general, have lower energy values, the docking energies are also lower, as compared to the molecule evolved using fixed size representation.