Computer-aided molecular design using the Signature molecular descriptor: Application to solvent selection

摘要

There is a growing demand to develop more environmentally friendly solvents to reduce costs and comply with regulation. Researchers at GlaxoSmithKline (GSK) have developed a solvent selection guide that ranks 47 frequently used solvents from 1 to 10 in five areas related to environmental compatibility. In this work, we apply a computer-aided molecular design method known as inverse design with the Signature molecular descriptor to identify additional potentially green solvents outside of GSK's list. Applying this approach is much quicker, less expensive and allows for a more comprehensive search for the most suitable candidates than working with experimental data alone. We present results for solvents with optimal predicted properties that span the classes from the 47 compounds in the GSK solvent selection guide and include several which are hybrids that cross-cut amongst classes. Additionally, our technique "rediscovers" the known green solvent ethyl lactate through this method by combining different solvent classes.