A new GRID-based method for scaffold hopping (SHOP) is presented. In a fully automatic manner, scaffoldswere identified in a database based on three types of 3D-descriptors. SHOP's ability to recover scaffoldswas assessed and validated by searching a database spiked with fragments of known ligands of three differentprotein targets relevant for drug discovery using a rational approach based on statistical experimental design.Five out of eight and seven out of eight thrombin scaffolds and all seven HIV protease scaffolds wererecovered within the top 10 and 31 out of 31 neuraminidase scaffolds were in the 31 top-ranked scaffolds.SHOP also identified new scaffolds with substantially different chemotypes from the queries. Docking analysisindicated that the new scaffolds would have similar binding modes to those of the respective query scaffoldsobserved in X-ray structures. The databases contained scaffolds from published combinatorial libraries toensure that identified scaffolds could be feasibly synthesized.