Date Published: August 01, 2016
Publisher: International Union of Crystallography
Author(s): Jason C. Cole, Colin R. Groom, Murray G. Read, Ilenia Giangreco, Patrick McCabe, Anthony M. Reilly, Gregory P. Shields.
An investigation into using shape-similarity of molecules to generate putative crystal structures.
Effective and efficient methods for crystal structure prediction (CSP) are much sought after (Price, 2014 ▸). Much of the focus of these methods has been directed towards identifying the thermodynamically most stable lattice a molecule can form. This is a formidable challenge. Firstly, one must identify plausible conformations of the molecule of interest. An enormous range of potential lattices consisting of these conformers must then be generated. These trial structures are then typically subjected to energy minimization and the resulting energies calculated to identify those with the lowest.
Generating structures based on pre-existing CSD structural analogues is useful for the generation of structural lattices. For a test set of 409 predominantly drug-like molecules, the method is successful in generating an analogue that minimizes to the same local minimum as the observed structure in 83.6% of cases, if the first 100 000 analogues suggested by shape similarity are considered.