Screening of drug databases

Please note that the “Correlation Vectors ” (CV) similarity measure referred to in this text corresponds to the CATS descriptor. Ligand Similarity-based virtual screening The experimental determination of protein structures is a difficult task and is not routinely possible. Especially membrane-bound proteins are difficult to crystallize [16]. Therefore, the situation that the target protein structure is unknown is common at an early stage of a pharmaceutical project. In such cases, only the structure of molecules already known to bind to the target protein, for example an endogenous ligand, can be used for virtual screening. When a molecule binds to the protein, it adopts a conformation that sterically and electrostati-cally fits into the active site. In order to detect compounds possibly binding into the same active site, one can search for molecules that can adopt conformations that are similar to the given active molecule. This similarity is defined with respect to size, shape and spatial distribution of functional groups and can be regarded as “chemical similarity ” [17-19]. In principle, however, we are interested in describing “biological similarity ” in the sense that molecules exhibiting the same biological response will be considered as similar. Biological similarity, however, depends o