Although databases such as PubChem contain 3D structures for millions of molecules, there is a much larger space of small drug-like molecules that researchers wish to explore. Next-generation drug discovery projects require a high-throughput way of predicting accurate 3D structures of virtual molecules. State-of-the-art density functional theory (DFT) methods are accurate but slow, while lower levels of theory such as molecular mechanics models are insufficient for describing the complex bonding of organometallic molecules. A recent system named COSMOS has demonstrated its ability to quickly predict 3D structures for millions or billions of virtual molecules given a large library of precomputed rigid fragments. Here we present an extensible library of 100,000 unique, highly accurate fragment structures that we have produced from isomeric SMILES codes, using a combination of pattern matching, molecular mechanics, and DFT.

High-throughput 3D structure prediction of small molecules

243rd National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.