MGAC: Frist Principle CSP

J. C. FACELLI; A. M. LUND; G. I. PAGOLA; A. M. ORENDT; M. B. FERRARO

Cambridge

Workshop; Sixth blind test of organic crystal-structure prediction methods; 2015

Cambridge Crystallographic Data Center (CCDC)

The predicted crystal structures submitted by participants were compared to the experimentally determined known crystal structures using the Crystal Packing Similarity Tool (J. Appl. Cryst., 2005, 38, 228-231), as used in previous blind tests.1 The tool represents a crystal structure using a cluster of N molecules comprised of a central reference molecule and (N-1) nearest-neighbour molecules. The distances and a subset of the triangles that define the reference cluster are then used as 3-D substructure-search query within the comparison structure. For this search two molecules are considered to match if these distances agree within 20% and the angles of the triangles agree within 20°. Those molecules that match are then overlaid and a root mean-squared deviation (RMSD) is calculated. The result of the comparison is a number of molecules that match, n, between the two packing shells and a corresponding RMSDn for those matching molecules.  Predicted structures were deemed to match an experimental structure when 20 out of 20 molecules matched with an RMSD20 of less than 1.0 Å, although in practice all matches were below the RMSD threshold. Where multiple clusters can be defined for an input crystal (i.e. Z' > 1 or structures submitted in P1 symmetry) the best result is retained. The Crystal Packing Similarity Tool normally considers only heavy atoms in its clustering-matching and RMSD analysis, ignoring H-atom positions due to their limited accuracy in standard X-ray crystal structures. Their presence or connection to heavy atoms is not ignored though.  For XXIII, some of the predicted crystal structures have the same heavy-atom positions as the experimental structure but place the carboxylic-acid H-atom on the oxygen closest to the NH group. The analysis for these systems was therefore performed twice, once requiring the H atom to be located as in the experimental structure and a second time where the Hatom location and connectivity was not considered.  In the case of XXIV, a cluster is only generated from the heavy component but each of the three components in the asymmetric unit counts towards N, and therefore a cluster of 20 components does not amount to the same physical extent as for the other systems. In addition, H-atom positions will be particularly important for this system. Therefore, initial analysis was performed ignoring H-atom positions and with N = 20. If a match was found, the analysis for that structure was re-run considering H-atom positions and with N = 60 to confirm the match.  Finally, in some predicted structures of XXV, proton transfer between the co-formers occurred, yielding a salt. While we are confident that the experimental structure is a cocrystal, with no proton transfer, checks were made to see if the iso-structural salt structure was predicted by any of the submissions.