Yvonne chose 11 speakers, including many of her colleagues from Abbott. The first talks were devoted to those who had the vision to collect and curate structural data. ACS has mounted these and some of the other presentations online as PowerPoint slides synched with audio [1]. First off was Frank Allen of the Cambridge Crystallographic Data Centre (CCDC) who discussed how smallmolecule crystal structure information aids drug discovery and development. He noted that Stahl and co-workers had recently presented a thorough review [2] of the applications of small-molecule crystal conformations in drug discovery. Despite the fact that higher-energy conformations can occur, eg, the presence of planar biphenyl conformers that are about 6 kJ/mol above the minimum energy twisted form [3], or small conformational variations due to favorable intermolecular interactions, the review showed that crystal conformations are very valuable experimental information in a modeling environment. Frank’s team had also investigated whether higherenergy conformers were a common occurrence in crystal structures [4]. They showed that torsion angles associated with higher strain energy ([5 kJ/mol) appear to be very unusual and, in general, high-energy conformers may well be under-represented in crystal structures when compared with a gas-phase, room-temperature Boltzmann distribution. Since crystallographically determined bond lengths, angles and torsions from the Cambridge Structural Database (CSD) are so useful, CCDC has produced a knowledge base, Mogul, to make them more readily accessible [5]. Validation experiments have shown that, with rare exceptions, such as those noted above, search results from Mogul give precise experimental information on molecular geometrical preferences.

Crystal structures are also the principal source of information on intermolecular interactions such as hydrogen bonding, dipole–dipole, and halogen–halogen interactions. CCDC’s knowledge base IsoStar [6] incorporating crystal structure data from the CSD and the Protein Data Bank (PDB), and interaction energies obtained using ab initio intermolecular perturbation theory (IMPT), enables users to visualize and study a wide range of interactions that are important in protein–ligand docking. Recent work at CCDC has included a thorough analysis of the energetic effects of varying the angle at the donor hydrogen in hydrogen bonds [7], together with combined CSD/IMPT studies of some less common interactions [8–11]. Frank cited a valuable review of interactions that are not mediated by hydrogen [12]. Finally Frank discussed the use of crystal structure information in the development, formulation and delivery of active pharmaceutical ingredients (APIs). An unexpected polymorph can have severe financial implications as exemplified in 1998, when supplies of Norvir capsules were suspended, when a new, much less soluble crystal form of the API ritonavir [13] suddenly appeared in production. CCDC has developed a method, based on a statistical analysis of hydrogen bonds in the CSD, to calculate the propensity for formation of different types of hydrogen bonds, given just the chemical diagram of a potential API. Applied now to ritonavir, the logit hydrogen-bonding propensity (LHP) model [14–16] showed that the original polymorphic form I of ritonavir was probably a metastable form, and that a more stable form was highly likely, as proved to be the case. In conclusion, Frank noted that ‘‘crystals are windows on the world of atoms’’[17], and