IntroductionMolecular recognition in biological systems relies on a sophisticated interplay of several nonbonded interactions between ligands and their binding sites in addition to other important factors like binding kinetics and desolvation. Insight into those interactions became possible due to the tremendous increase of 3D structural information of protein targets for medical therapy, as exemplified by the growth of the PDB database (RCSB [1,2]) with now ∼70 000 entries (December 2010). Detailed database analysis provides many structurally characterised examples for ligands favourably interacting with their target protein and thus allows studying the relevant interaction motifs contributing to the experimentally observed binding affinity. In addition, guidelines on favourable motifs can be deduced based on interaction geometries and associated affinity data [3].In order to identify lead structures for optimisation, biophysical and in silico lead finding techniques, like virtual screening [4][5][6][7] or structure-based design, are integral components in the industrial drug-discovery setting today and complementing the routinely applied highthroughput screening (HTS) [8,9]. Capitalising on 3D structural knowledge, computational approaches are cost effective and powerful, and can potentially reduce the number of in-vitro assays after in-silico selection to a few hundreds or thousands.