State-of-the-Art in Ligand-Based Virtual Screening

Posted by

SELECTBIO

Computational approaches of different sophistication are successfully employed in the search for novel active compounds. Among these are various holistic similarity methods that often utilize only simple molecular representations. It is evident that the complexity of computational tools and molecular representations does not correlate with their success in virtual screening. Why is this so? Why do many different virtual screening methodologies frequently display comparably good or poor performance on given compound classes? The inability to rationalize virtual screening performance and predict successes or failures presents one of the grand challenges in this field. However, systematic analyses of structure-activity relationships of compound classes directed against different targets offer some answers to these still open questions. Considering that much of the virtual screening work carried out in pharmaceutical settings is not disclosed, and that generally only successful applications are published, it is difficult to draw firm conclusions about the current state of this field. At least, a comprehensive literature survey of prospective virtual screening applications is available that mirrors the state-of-the-art and also highlights a number of scientific issues that frequently limit the quality of virtual screening investigations. Furthermore, exemplary virtual screening case studies addressing different types of targets are presented that illustrate opportunities and limitations of contemporary ligand similarity-based methods.