iSYN: de novo drug design with click chemistry support

Abstract

We present iSyn, an evolutionary algorithm that automatically designs de novo ligands with high predicted binding affinity and drug-like properties. It attempts to optimize candidate ligands in accordance with click chemistry and thus ensures chemical synthesizability. In addition to the existing genetic operators of mutation and crossover inherited from AutoGrow 3.0, our iSyn introduces four novel genetic operators to "cut" ligands in order to prevent them from becoming too large in molecular size, hence preserving drug-like properties. Moreover, iSyn interfaces with our fast docking engine idock, greatly reducing the execution time. We hope iSyn can supplement medicinal chemists' efforts.

iSyn was applied to optimizing candidate ligands against two important drug targets, TbREL1 and HIV-1 RT, and managed to produce chemically valid ligands with high predicted binding affinities and drug-like properties. In the example of TbREL1, the predicted free energy of the best generated ligand decreased from -9.878 kcal/mol to -13.985 kcal/mol after 3 generations. In the example of HIV-1 RT, the predicted free energy of the best generated ligand decreased from -5.427 kcal/mol to -12.488 kcal/mol after 2 generations, meanwhile the molecular mass dropped from 602.818 Da to 461.736 Da, so that the compound could be properly absorbed by human body.

iSyn is written in C++ and Python, and is free and open source, available at http://istar.cse.cuhk.edu.hk/iSyn.tgz. It has been tested successfully on Linux and Windows. In the near future we plan to implement a web-based user interface to facilitate its usage and to promote large-scale de novo drug design.