Protein evolution has employed a small repertoire of a few thousand elementary modules or domain families. Present-day proteins are formed from a series of domains, each of which can be assigned to one of these ancient families. Since the structure and function within a family are largely conserved, identification of domains has become a powerful tool in protein sequence analysis and an active research area. In the talk I will describe our group's effort to produce alignments that represent the conserved 3-dimensional core structure of protein domain families. Our goal is to identify as precisely as possible the residues forming this core structure and any binding or catalytic sites that derive from it. I will also describe CD-Search (Conserved Domain Search), a sequence-similarity search service for identification of conserved domains. This service is available at http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml.