The interactions between proteins and other molecules are critical to a lot of biological processes, e.g. cellular signaling, regulation of cell cycles. Due to the fact that the functions of proteins are not necessarily correlated with protein sequence or fold, local feature analysis of proteins especially the pattern of binding sites may be another way to further understand protein functions and structures. In this study, a 3D structure based evaluation method for exploring the binding site similarity named PocketShape is presented. By means of structure-based alignment, PocketShape is capable of detecting common patterns or residue reservation between binding sites which are not required to possess continuous residues or to be homologous. Further possible application of PocketShape includes: prediction of a drug candidate’s off-target interactions, prediction of new target for known drugs, prediction of a protein’s function, protein classification. A protein profiling test carried out on protein kinase dataset demonstrates that PocketShape is capable of retrieving binding sites with common pattern.