HYDROGEN BONDS AND OTHER NON-COVALENT INTERACTIONS: DATABASE ANALYSES, AB INITIO MO STUDIES AND KNOWLEDGE BASES
Frank H. Allen
Cambridge Crystallographic Data Centre 12 Union Road, Cambridge CB2 1EZ, England
Keywords: Hydrogen Bonding, Database Analyses, Cambridge Structural Database, Protein Data Bank, Ab Initio MO calculations, Intermolecular Perturbation Theory, Intermolecular Interactions IsoStar Knowledge Base, Oxygen...Halogen Interactions, Carbonyl...Carbonyl Interactions Hydrogen Bonded Motifs
Systematic knowledge of the types, geometries and relative strengths of non-covalent interactions is crucial in many areas of supramolecular research, e.g. crystal engineering, crystal structure prediction, rational drug design, protein - ligand docking, etc. Crystal structures, as archetypal supermolecules, identify important non-covalent bonds and provide direct experimental observations of interaction geometries.
The crystallographic databases, particularly the Cambridge Structural Database (CSD), have always been heavily used in systematic studies of hydrogen bonding. In recent years, these studies have diversified from the ubiquitous and strong H-bonds formed between N-H and O-H donors and oxygen and nitrogen acceptors to studies of more specific phenomena and, particularly, to the characterisation of weaker H-bonds. Thus, we may list:
Studies of this type are immensely valuable, but can only provide rough estimates of the relative strengths of interactions. Thus it is important to supplement the crystallographic results with high-level ab initio molecular orbital calculations of interaction energies for selected model systems.
In recent years, we have combined systematic database analyses with calculations that use the intermolecular perturbation theory (IMPT) of Hayes and Stone [6] applied to model dimers. The technique is free of basis set superposition errors and generates an interaction energy that is the summation of five components. Since IMPT calculations are cpu-intensive, our approach is to use CSD surveys to identify preferred starting geometries for the examination of local features of the PE hypersurface.
The combined CSD-IMPT approach has been applied to a number of H-bonding situations and has also been extended to the examination of intermolecular interactions that are not mediated by hydrogen, e.g. oxygen....halogen [7] and carbonyl...carbonyl [8] interactions.
The existence of specific and robust intermolecular interaction motifs, designated as 'supramolecular synthons' by Desiraju [9], is of crucial importance in crystal engineering, and many of these motifs involve hydrogen bonding. Recently [10] we have used the CSD to provide an automatic ranking of H-bonded motifs on the basis of their probabilities of formation. This work now also generates graph-set descriptors [11] of H-bonded systems and provides a basis for improved graphical visualisations of H-bonded networks.
Given the importance of knowledge concerning non-covalent bonding that can be generated (relatively) routinely, the CCDC is now compiling and distributing IsoStar [12]. This knowledge-based library of intermolecular interactions is generated from the CSD, the PDB and from selected IMPT calculations. IsoStar will be noted here, but is more fully described elsewhere at this Meeting [V.J. Hoy, Microsymposium F3: Structural Databases].