Development and testing of a de novo drug-design algorithm.

In this article we present an implementation of a de novo drug-design algorithm. The algorithm starts with a molecule placed in the binding site of a protein and then modifies it using a sequential growth approach. This involves successive cycles of suppression of randomly picked groups in the molecule and their replacement by other groups chosen from databanks of linear or cyclic fragments. The algorithm has been coupled with the DYNAMO library which allows the simulation of macromolecules using molecular mechanical and quantum chemical methods. The main body of the article describes the methodologies we use to create, characterize and evaluate putative ligands. We also consider briefly an application of the algorithm to a protein of pharmacological interest, the neuraminidase of the influenza virus, and discuss the strengths and weaknesses of our approach.