Jos P. M. Lommerse

Crystal structure prediction of small organic molecules: a second blind test.

The first collaborative workshop on crystal structure prediction (CSP1999) has been followed by a second workshop (CSP2001) held at the Cambridge Crystallographic Data Centre. The 17 participants were given only the chemical diagram for three organic molecules and were invited to test their prediction programs within a range of named common space groups. Several different computer programs were used, using the methodology wherein a molecular model is used to construct theoretical crystal structures in given space groups, and prediction is usually based on the minimum calculated lattice energy. A maximum of three predictions were allowed per molecule. The results showed two correct predictions for the first molecule, four for the second molecule and none for the third molecule (which had torsional flexibility). The correct structure was often present in the sorted low-energy lists from the participants but at a ranking position greater than three. The use of non-indexed powder diffraction data was investigated in a secondary test, after completion of the ab initio submissions. Although no one method can be said to be completely reliable, this workshop gives an objective measure of the success and failure of current methodologies.

Carbonyl-carbonyl interactions can be competitive with hydrogen bonds

The geometries and attractive energies of carbonyl–carbonyl interactions have been investigated using crystallographic data and ab initio molecular-orbital calculations. Analysis of crystallographic data for 9049 carbon-substituted >C=O groups shows that 1328 (15%) form contacts with other >C=O groups, in which d(C⋯O) < 3.6 A. Three common interaction motifs are observed in crystal structures: (a) a slightly sheared antiparallel motif (650 instances) involving a pair of short C⋯O interactions, together with (b) a perpendicular motif (116 instances) and (c) a highly sheared parallel motif (130 instances), which both involve a single short C⋯O interaction. Together, these motifs account for 945 (71%) of the observed interactions. Ab-initio-based molecular-orbital calculations (6-31G** basis sets), using intermolecular perturbation theory (IMPT) applied to a bis-propanone dimer model, yield an attractive interaction energy of −22.3 kJ mol−1 for a perfect rectangular antiparallel dimer having both d(C⋯O) = 3.02 A and attractive energies < −20 kJ mol−1 over the d(C⋯O) range 2.92–3.32 A. These energies are comparable to those of medium-strength hydrogen bonds. The IMPT calculations indicate a slight shearing of the antiparallel motif with increasing d(C⋯O). For the perpendicular motif, IMPT yields an attractive interaction energy of −7.6 kJ mol−1, comparable in strength to a C—H⋯O hydrogen bond and with the single d(C⋯O) again at 3.02 A.

IsoStar: A library of information about nonbonded interactions

Crystallographic and theoretical (ab initio) data on intermolecular nonbondedinteractions have been gathered together in a computerised library(’IsoStar‘). The library contains information about the nonbonded contactsformed by some 250 chemical groupings. The data can be displayed visually andused to aid protein–ligand docking or the identification of bioisostericreplacements. Data from the library show that there is great variability inthe geometrical preferences of different types of hydrogen bonds, although ingeneral there is a tendency for H-bonds to form along lone-pair directions.The H-bond acceptor abilities of oxygen and sulphur atoms are highly dependenton intramolecular environments. The nonbonded contacts formed by manyhydrophobic groups show surprisingly strong directional preferences. Manyunusual nonbonded interactions are to be found in the library and are ofpotential value for designing novel biologically active molecules.

Hydrogen bonding of carbonyl, ether, and ester oxygen atoms with alkanol hydroxyl groups

An attractive way to study intermolecular hydrogen bonding is to combine analysis of experimental crystallographic data with ab initio—based energy calculations. Using the Cambridge Structural Database (CSD), a distributed multipole analysis (DMA)‐based description of the electrostatic energy, and intermolecular perturbation theory (IMPT) calculations, hydrogen bonding between donor alkanol hydroxyl groups and oxygen acceptor atoms in ketone, ether, and ester functional groups is characterized. The presence and absence of lone pair directionality to carbonyl and ether or ester oxygens, respectively, can be explained in terms of favored electrostatic energies, the major attractive contribution in hydrogen bonding. A hydrogen bond in its optimum geometry is only slightly stronger when formed to a ketone group than to an ether group. Hydrogen bonds formed to carbonyl groups have similar properties in a ketone or ester, but the ester O2 differs from an ether oxygen due to various environmental effects rather than a change in its intrinsic properties. For (E)‐ester oxygens, there are few hydrogen bonds found in the CSD because of the competition with the adjacent carbonyl group, but the interaction energies are similar to an ether. Hydrogen bonds to O2 of (Z)‐esters are destabilized by the repulsive electrostatic interaction with the carbonyl group. The relative abundance of nonlinear hydrogen bonds found in the CSD can be explained by geometrical factors, and is also due to environmental effects producing slightly stronger intermolecular interaction energies for an off‐linear geometry.

Hydrogen bonding properties of oxygen and nitrogen acceptors in aromatic heterocycles

The directionality and relative strengths of hydrogen bonds to monocyclic aromatic heterocycles were investigated using crystal structure data and theoretical calculations. Surveys of the Cambridge Structural Database for hydrogen bonds between C(sp3)(SINGLE BOND)O(SINGLE BOND)H and aromatic fragments containing one or more nitrogen and/or oxygen heteroatoms showed that hydrogen bonds to nitrogen atoms are much more abundant than to oxygen. Distinct preferred orientations were also revealed in these surveys. Theoretical calculations were performed on the interaction of methanol with pyridine, pyrimidine, pyrazine, pyridazine, oxazole, isoxazole, 1,2,4‐oxadiazole, and furan as models for the heterocyclic fragments. The intermolecular potential surface was thoroughly scanned using a model potential that accurately described the electrostatic forces (derived from distributed multipole analysis) with empirical parameters for the repulsion and dispersion terms. Minima on this surface agreed well with the observed orientations in the data base and they were typically deeper for nitrogen than for oxygen acceptors, although the hydrogen bond strength and geometry was influenced by other heteroatoms in the ring. These results were confirmed by highly accurate intermolecular perturbation theory calculations, which also estimated the deviations from hydrogen bonding in the traditional nitrogen lone pair direction that could occur with negligible reduction in the interaction energy. © 1997 John Wiley & Sons, Inc. J Comput Chem 18: 2060–2074, 1997

Hydrogen-bond acceptor properties of nitro-O atoms : A combined crystallographic database and Ab initio molecular orbital study

Crystallographic data for 620 C-nitro-O...H-N,O hydrogen bonds, involving 560 unique H atoms, have been investigated to the van der Waals limit of 2.62 A. The overall mean nitro-O...H bond length is 2.30 (1) A, which is much longer (weaker) than comparable hydrogen bonds involving >C=O acceptors in ketones, carboxylic acids and amides. The donor hydrogen prefers to approach the nitro-O atoms in the C-NO 2 plane and there is an approximate 3:2 preference for hydrogen approach between the two nitro-O atoms, rather than between the C and O substituents. However, hydrogen approach between the two O acceptors is usually strongly asymmetric, the H atom being more closely associated with one of the O atoms: only 60 H atoms have both O...H distances ≤ 2.62 A. The approach of hydrogen along putative O-atom lone-pair directions is clearly observed. Ab-initio-based molecular orbital calculations (6-31G ** basis set level), using intermolecular perturbation theory (IMPT) applied to the nitromethane-methanol model dimer, agree with the experimental observations. IMPT calculations yield an attractive hydrogen-bond energy of ca -15 kJ mol -1 , about half as strong as the >C=O...H bonds noted above.

Characterising Non-Covalent Interactions With The Cambridge Structural Database

This review describes how the CSD can be used to study non-covalent interactions. Several different types of information may be obtained. First, the relative frequencies of various interactions can be studied; for example, we have shown that the terminal oxygen atoms of phosphate groups accept hydrogen bonds far more often than the linkage oxygens. Secondly, information can be obtained about the geometries of nonbonded contacts; for example, hydrogen bonds to P-O groups rarely form along the extension of the P-O bond, whereas short contacts between oxygen and carbon-bound iodine show a strong preference for linear C-I ... O angles. Thirdly, the CSD can be searched for novel interactions which may be exploited in inhibitor design; for example, the I ... O contacts just mentioned, and N-H ... pi hydrogen bonds. Finally, the CSD can suggest synthetic targets for medicinal chemistry; for example, molecules containing delocalised electron deficient groups such as trimethylammonium, pyridinium, thaizolium and dinitrophenyl have a good chance of binding to an active-site tryptophan. Although the CSD contains small-molecule crystal structures, not protein-ligand complexes, there is considerable evidence that the contacts seen in the two types of structures are similar. We have illustrated this a number of times in the present review and additional evidence has been given previously by Klebe. The major advantages of the CSD are its size, diversity and experimental accuracy. For these reasons, it is a useful tool for modellers engaged in rational inhibitor design.

Use of the Cambridge Structural Database to Study Non-Covalent Interactions: Towards a Knowledge Base of Intermolecular Interactions

A knowledge of the preferred non-covalent interaction modes of molecules, through their functional substructures, is important to the design of novel bioactive molecules. Such knowledge permits us to model the interactions of putative bioactives at known binding sites, or to make inferences about binding site structure from the structures of known bioactives. Thus a protein-ligand complex can be described as an assembly of covalently bonded units or ions that is organised according to the diverse weak forces that govern non-covalent interactions: a phrase that accurately defines a supermolecule [1].