Cavka – a new automatic pharmacophore elucidation method in progress

Three dimensional pharmacophore models can be considered as an ensemble of steric and electronic features in space, which are necessary to ensure intermolecular interaction with a specific target in order to trigger or to block biological activity [1]. By identifying these features, a 3D pharmacophore model can be built in order to screen multi-conformatorial databases with the aim to detect compounds matching the pharmacophoric hypothesis and subsequently submit them to a biological testing. Even if a 3D crystal structure is at hand, the creation of a reliable pharmacophore model remains a challenging task. CavKA (Cavity Knowledge Acceleration), our own inhouse strategy employs the information of Co-crystallised ligand-receptor complexes for an automatic pharmacophore creation. Ligand features interacting with the binding site are detected and Grid [2] force field information is additionally taken into account as to weight and prioritize the identified features in question, to transform them into a pharmacophore model without any user intervention. Our method is compared to LigandScout [3] and a custom MOE [4] implementation, similar to LigandScout, two powerful standard tools. Both are identifying ligand-receptor interactions to highlight important ligand features to be selected for creating pharmacophore models automatically. The performance is evaluated in a retrospective screening on the FieldScreen [5] dataset outlining strengths, weaknesses and as well as similarities of each method for the scrutinized targets.