J.A.: Evolving molecules for drug design using genetic algorithms via molecular trees

We present a new representation for a genetic algorithm to evolve molecular structures representing possible drugs that bind to a given protein target receptor. Our representation is tree-structured with functional groups for leaves, and captures chemically relevant information efficiently. We assume a given target protein structure with known essential residues, and derive the placement of the functional groups in each chromosome from both lengths and the position of a pharmacore in the receptor. Our fitness evaluation takes into consideration both proximities and polarities of the functional groups of the evolved drug structure and the residues. Our evolved structures were intriguingly similar to known active anti-viral drug structures. Our experiments indicate that a tree-structured molecular representation and a simple evolutionary computation can design acceptable molecular structures that are potentially useful for drug design endeavors.