Computational methods for scaffold hopping

It is not uncommon in drug discovery that the core fragment, typically called scaffold, of a molecule with an interesting biological activity cannot be developed further because of issues related to intellectual property, physicochemical properties, metabolic stability, or toxicity, to name only a few reasons. In this situation, it is desirable to replace this molecule with another having a different chemical connectivity, but similar shape and pharmacophore features enabling it to interact in the same way with the target as the original molecule. Such a replacement is called scaffold hopping. Several ligand‐based virtual screening and scoring methods supporting the identification of novel ligands starting from known ligands and, if known, their bound conformation, are available. Scaffold hopping capability has been demonstrated for pharmacophore searches, field‐ and shape‐based similarity searches, alignment free similarity searches using three‐dimensional (3D)‐ or connectivity‐based descriptors, and fragment‐based methods. Although for many methods successful prospective uses have been reported, rigorous systematic benchmarking of scaffold hopping is still challenging due to the lacking consensus in the definition of a scaffold. Despite their drawbacks, computational scaffold extraction methods have been frequently used in approximate benchmarks for scaffold hopping. In many systematic retrospective studies, connectivity‐based methods were shown to be at least equally effective as 3D methods, especially when the conformations of the reference structures were generated with a conformer generator. However, in prospective comparison studies, especially when a hypothesis of the 3D binding conformation was available and used, often 3D methods were found to give superior results. © 2012 John Wiley & Sons, Ltd. This article is categorized under: Computer and Information Science > Chemoinformatics