Investigating crystal engineering principles using a data set of 50 pharmaceutical cocrystals Peter A. Wood, Cambridge Crystallographic Data Centre 24 th March 2009 ACS Spring Meeting, Salt Lake City, USA 1
Co-crystals The term co-crystal refers to a crystalline material that contains two or more neutral, organic molecules or a neutral molecule plus a salt In the context of pharmaceutical co-crystals, the molecule of most interest is the Active Pharmaceutical Ingredient (API) The additional molecule in a multi-component structure is referred to as the co-former 2
Co-crystals Co-crystals are slowly becoming more popular within the pharmaceutical industry Potentially greater flexibility and diversity than salts or hydrates To use co-crystals effectively, we need to understand & have some control over them A large family of co-crystals with a consistent host molecule is invaluable in learning about packing effects, motifs, H-bond competition and so on 3
Aims First multi-technique analysis of a large family of organic co-crystals Wanted to analyse structures using a range of methods to learn as much as possible Trying to avoid focussing solely on hydrogen bonds allow other features to show importance What can we learn about co-crystal design in general? 4
Carbamazepine Has anti-epileptic and analgesic properties Known to form at least four polymorphs Large family of multicomponent structures UK crystallography s favourite compound? Carbamazepine (CBZ) 5
Developing the dataset 37 structures retrieved from the CSD 13 new cocrystals determined through advanced co-crystal screening studies only carboxylic acid co-formers were screened In total, 50 CBZ structures Childs, Rodríguez-Hornedo et al. (2008) CrystEngComm, 10, 856-864. 6
Analytical tools New crystal form analysis tools in Materials module of the CSD visualiser (Mercury) arising from collaboration with Pfizer Institute for Pharmaceutical Materials Science: Hydrogen-bond motif analysis General packing feature analysis Crystal packing similarity analysis 7
Hydrogen-bond motif analysis Designed to provide search tools for interaction motifs Auto-generated sets of motifs Frequency of occurrence data (CSD) indicates how common or likely a motif is 37.7% 1.5% 22.3% 8
General packing feature analysis Easy selection of feature Simple searching to assess how unusual is this feature? & has this been seen before? Can relate structural features to stability 9
Packing similarity analysis Materials module similarity calculation Identifies clusters in common between crystal structures Can group structures based on level of geometric similarity 10
Aspects of analysis Start with description and analysis of hydrogen bonding interactions within the dataset The next talk (by Scott Childs, Renovo Research) will focus more on molecular shape-based packing features 11
Hydrogen bonding motifs Most obvious and well documented motif in CBZ structures is the carboxamide homodimer (a) occurs in all four polymorphs Carboxylic acid can provide competition the hetero-dimer (b) 12
Hydrogen bonding motifs Motif percentage frequencies observed in the CSD: Structures containing both groups Similar results seen in CBZ dataset for structures containing both groups (27% homo, 65% hetero) CBZ homo-dimer only perturbed by COOH group 33% 56% 3% 13
Hydrogen bonding motifs Can also use the motif analysis tools to investigate the more complicated motifs present In the analysis of co-crystals, the arrangement of the components in a motif is of interest Identified ring and chain patterns occurring as well as order of the groups interacting, i.e. CBZ or CF (coformer contact group) e.g. an infinite 2 molecule chain could be CBZ-CBZ-, CBZ-CF-, or CF-CF- 14
Rings R3 R4A R4B R6A R6B R6C R6D R6E 6 x x x 11 x 16 x x 17 x x x 18 x 19 x 20 x 21 x 23 x x x 24 x x 27 x x x 28 x x x 30 x 31 x x x x 38 x 43 x 45 x 50 x CBZ-CF-CBZ-CF- CBZ-CF- CBZ-CF-CBZ-CBZ- CF-CBZ- CBZ-CBZ-CF-CF- CBZ-CF-CBZ-CBZ- CF-CF- CBZ-CF-CF-CBZ-CF- CF- CBZ-CBZ-CF-CF-CF- CF- CBZ-CF-CBZ-CF- CBZ-CBZ-CF- 15 Hydrogen bonding motifs C1A C1B C2A C2B C3A C3B Infinite Chains 6 x x 8 x Motifs: 9 x 17 x x 21 x 23 x 24 x 27 x x Intermolecular CBZ CBZ contacts are not common 28 x x 32 x 37 x 38 x 39 x x 43 x 50 x CBZ-CBZ-CF- CF-CF-CBZ- CF-CBZ- CF-CF- CF-CF- CBZ-CBZ-
Hydrogen bonding motifs Role of co-former as a link: 16
Hydrogen bonding motifs Dimensionality: H-bond dimension # of structures % of structures Ave. # of Donors Ave. # of Acceptors 0D 17 42.5 0.8 1.5 1D 11 27.5 1 1.8 2D 7 17.5 2.6 2.1 3D 5 12.5 2.8 2.6 Preference for discrete aggregates High correlation between dimension of H-bonding network & average number of acceptors 17
Hydrogen bonding Lack of acceptor for second carboxamide N-H Structure 36 CBZ/benzoic acid (1:1) 18
Etter s first rule Etter s rules: How many CBZ structures violate the first rule? Is this unusual? Why does it happen? 19
Etter s first rule Start with evaluation of the CSD to get a broad perspective first How do you define a hydrogen bond? Selected very broad limits Q A H Q B [where Q A = N,O & Q B = N, O, S, F, Cl, Br, I] Contact defined as within sum of vdw radii + 0.3 Å 20
Etter s first rule Transformed variables to: normalise distances use spherical coordinates, solving angle bias problems Lommerse et al. (1996) JACS, 118, 3108-3116. 21
Etter s first rule Clear peak at θ = 180 Smaller peaks at 130 and 105 due to intramolecular contacts Peak finishes before x = 1 22
Etter s first rule 97.5% of strong hydrogen-bond donor hydrogen atoms have an acceptor within sum of vdw plus 0.3 Å with D-H A angle 90 In the CBZ dataset, 12 of the structures (24%) do not have such an acceptor Appears likely to be due to steric hindrance 23
Conclusions Developed a dataset of 50 structures containing the same pharmaceutically-relevant compound Applying newly developed analytical tools to learn more about co-crystal design Motif competition with the carboxamide dimer only seen to be successful using carboxylic acids Etter s first rule broken for 24% of the dataset indication that molecular shape is crucial 24
Acknowledgements Scott Childs (formerly with SSCI/Aptuit, now CEO of Renovo Research) Naír Rodríguez-Hornedo & L. Sreenivas Reddy (University of Michigan) Kenneth Hardcastle (Emory University) CCDC Elna Pidcock László Fábián James Chisholm Clare Macrae Funding: 25