DBD2BS uses the structure of DNA binding domains (DBDs) to predict the corresponding binding sites on DNA. Queries can be specified by a PDB ID or the coordinates of a DNA-binding protein in PDB format. Users do not have to mark the DBD, which will be identified by DBD2BS automatically. DBD2BS is the first web server that can predict binding sites from protein unbound structures, in addition to using native protein-DNA complexes directly. For the users’ convenience, DBD2BS currently maintains a list of 2,176 protein-DNA complexes and 3,135 unbound structures of DNA-binding proteins collected from PDB.